state of the art in simulation-based optimisation for maintenance systems

8/18/2019 State of the Art in Simulation-based Optimisation for Maintenance Systems

1/16

Survey

State of the art in simulation-based optimisation for maintenancesystems q

Abdullah Alrabghi ⇑ , Ashutosh TiwariManufacturing Department, Craneld University, Craneld, United Kingdom

a r t i c l e i n f o

Article history:Received 10 June 2014Received in revised form 18 December 2014Accepted 21 December 2014Available online 29 December 2014

Keywords:SimulationOptimisationMaintenanceModellingReview

a b s t r a c t

Recently, more attention has been directed towards improving and optimising maintenance in manufac-turing systems using simulation. This paper aims to report the state of the art in simulation-basedoptimisation of maintenance by systematically classifying the published literature and outlining maintrends in modelling and optimising maintenance systems. The authors investigate application areasand published real case studies as well as researched maintenance strategies and policies. Much of theresearch in this area is focusing on preventive maintenance and optimising preventive maintenancefrequency that will lead to the minimum cost. Discrete event simulation was the most reported techniqueto model maintenance systems whereas modern optimisation methods such as Genetic Algorithms wasthe most reported optimisation method in the literature. On this basis, the paper identies the currentgaps and discusses future prospects. Further research can be done to develop a framework that guidesthe experimenting process with different maintenance strategies and policies. More real case studiescan be conducted on multi-objective optimisation and condition based maintenance especially in aproduction context.

2014 Elsevier Ltd. All rights reserved.

1. Introduction

Maintenance aims to combat the inevitable degradation of assets over their operational lifetime and keep them in a workingorder. Therefore, maintenance plays an important role insustaining and improving assets availability, which in turn affectsthe productivity of the system in interest.

Recently, more attention has been directed towards improvingand optimising maintenance in manufacturing systems.Maintenance cost can reach anywhere between 15% and 70% of production costs (Wang, Chu, & Mao, 2008). Wang (2008) observesthat there is still a large potential for increasing the productivity incurrent maintenance practices. In some industries, a slightimprovement in throughput could result in signicant economicimpact (Yao, Fernández-Gaucherand, Fu, & Marcus, 2004).

Simulation delivers an advantage over analytical approachesbecause many maintenance policies are not analytically traceable(Nicolai & Dekker, 2008). In addition, it allows experimentingand better understanding of complex systems (Sebastian, 2006).

The complexity of maintenance systems has increased signi-cantly (Duffuaa, Ben-Daya, Al-Sultan, & Andijani, 2001; Nicolai &Dekker, 2008). This is partly due to modern manufacturing systemswhich involve numerous interactions and dependencies betweencomponents. It is evident that analytical and mathematicalapproaches are limited in solving such complex maintenance prob-lems. By developing both analytical and simulation models to solvethe same problem, Rezg, Chelbi, and Xie (2005) found that itresulted in a complex analytical model with unrealistic assump-tions compared to the simulation model which provided more ex-ibility and simpler estimations. Several studies have indicated thepreference of simulation to optimise maintenance problems overanalytical and mathematical approaches (Allaoui & Artiba, 2004;Briš, 2008; Gupta & Lawsirirat, 2006; Marseguerra, Zio, &Podollini, 2002).

Although research on maintenance optimisation was estab-lished decades ago (Dekker, 1996), the area of simulation-basedoptimisation in maintenance is becoming an emerging trend(Garg & Deshmukh, 2006; Sharma, Yadava, & Deshmukh, 2011).Simulation has been traditionally used as a tool to understand

http://dx.doi.org/10.1016/j.cie.2014.12.022

0360-8352/ 2014 Elsevier Ltd. All rights reserved.

Abbreviations: CBM, Condition Based Maintenance; CM, Corrective Mainte-nance; DES, Discrete Event Simulation; GA, Genetic Algorithms; PM, PreventiveMaintenance; SA, Simulated Annealing.

q This manuscript was processed by Area Editor Min Xie.⇑ Corresponding author.

E-mail addresses: [email protected] (A. Alrabghi), [email protected] (A. Tiwari).

Computers & Industrial Engineering 82 (2015) 167–182

Contents lists available at ScienceDirect

Computers & Industrial Engineering

j ou rna l ho m epag e : www.e l s ev i e r. com/ loca t e / ca i e

http://dx.doi.org/10.1016/j.cie.2014.12.022mailto:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1016/j.cie.2014.12.022http://www.sciencedirect.com/science/journal/03608352http://www.elsevier.com/locate/caiehttp://www.elsevier.com/locate/caiehttp://www.sciencedirect.com/science/journal/03608352http://dx.doi.org/10.1016/j.cie.2014.12.022mailto:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1016/j.cie.2014.12.022http://-/?-http://-/?-http://-/?-http://-/?-http://crossmark.crossref.org/dialog/?doi=10.1016/j.cie.2014.12.022&domain=pdfhttp://-/?-http://-/?-


2/16

and experiment with a system. However, connecting the simula-tion model to an optimisation engine ensures better and fasterresults. As illustrated in Fig. 1-1, simulation based optimisation isan approach whereby an optimisation engine provides the decisionvariables for the simulation program. The simulation program willrun the model and provide the results of the optimisation objectivefunction. This process will continue iteratively between the simu-lation program and the optimisation engine until it results in a sat-isfactory solution or a termination due to prescribed conditions(Amaran, Sahinidis, Sharda, & Bury, 2014).

A considerable amount of literature has been published onmaintenance simulation and optimisation. Dekker (1996) provideda comprehensive view and analysis of maintenance optimisationmodels and applications. It is interesting to note that in his work,simulation has not been mentioned and the emphasis was onmathematical models only. More recently, Sharma et al. (2011)observed that there is a potential as well as a growing interestamongst researchers to utilise simulation in optimising mainte-nance systems. The advancement in technology has enabledresearches to use powerful computers and software with decreas-ing costs. Vasili, Hong, Ismail, and Vasili (2011) review highlightedthat it is becoming increasingly difcult to rely on static solutiontechniques to optimise maintenance systems and ignore thedynamic and stochastic nature of current business environments.

On the other hand, Andijani and Duffuaa (2002) evaluated sim-ulation studies in maintenance systems in terms of adherence tosound modelling principles such as program verication and vali-dation. Alabdulkarim, Ball, and Tiwari (2013) reviewed the applica-tions of simulation in maintenance systems and categorised itaccording to the purpose of the study. Their research conrms thatresearch on maintenance simulation is steadily rising. Additionally,they observed that research on the combined use of simulation andoptimisation is limited.

Thus, this study provides an exciting opportunity to advanceour knowledge on the state of art in the combined use of simula-tion and optimisation in maintenance systems. It is complemen-tary to the earlier work of Alrabghi and Tiwari (2013) whichprovided a summary description of the available literature on thesame subject.

2. Review methodology

This research aims to identify and summarise available litera-ture on simulation-based optimisation of maintenance operations.Thus the scope is focused on research that includes simulatingmaintenance systems and connecting the simulation model to anoptimisation engine.

Research that focus on improving maintainability and reliabilityat the design stage is disregarded. There have been attempts tosimulate maintenance operations through static system models,usually using Monte-Carlo simulation (Marseguerra et al., 2002;Wang et al., 2008). As time is a signicant variable in maintenanceoperations, only attempts to model it through dynamic systemmodels are within the scope of this research.

The research methodology of this review is similar to that of Alrabghi and Tiwari (2013). However, the search in this review ismore comprehensive as it covers abstracts and keywords in addi-tion to titles. Moreover, search in conference papers was limitedto the last ten years only to capture recent work that is yet to bepublished in peer-reviewed journal papers.

A systematic research was conducted by searching for the fol-lowing keywords in article titles, abstracts and keywords: (main-tain ⁄ and optim⁄ and simulat⁄ ) and (maintenance and optim⁄and simulat ⁄ ). Scopus and Web of Science citation databases, twoof the largest abstract and citation databases of peer-reviewed lit-erature, were searched to identify the targeted papers. The Scopussearch resulted in 15,001 documents in English whereas the Webof Science search resulted in 9132 documents in English. An over-view of the review methodology is shown in the gure below: (seeFig. 2-1)

The resulting documents were ltered in a systematic method-ology as follows:

Excluding irrelevant subject areas such as medicine, social sci-ences and arts and humanities. The main relevant subject areasare engineering, mathematics, decision sciences and businessmanagement.

Reviewing the titles and abstracts. This includes reading titlesand abstracts and excluding papers that do not include simula-tion optimisation in maintenance.

Skimming the whole paper to nd out the application area aswell as optimisation methods and simulation techniques. Thiswas usually obtained by reading the methodology section of the paper.

A further comprehensive reading was conducted through thefull documents which yielded 59 articles after removing duplica-tions (Ali, Chen, Yang, Lee, & Jun, 2008; Allaoui & Artiba, 2004;

Fig. 1-1. Simulation based optimisation approach.

Fig. 2-1. Systematic review methodology.

168 A. Alrabghi, A. Tiwari / Computers & Industrial Engineering 82 (2015) 167–182

http://-/?-


3/16

Alrabghi, Tiwari, & Alabdulkarim, 2013; Arab, Ismail, & Lee, 2013;Asadzadeh & Azadeh, 2014; Boschian, Rezg, & Chelbi, 2009;Boulet, Gharbi, & Kenn, 2009; Cavory, Dupas, & Goncalves, 2001;Chen, Hsu, & Chen, 2006; Cheu, Wang, & Fwa, 2004; Chootinan,Chen, Horrocks, & Bolling, 2006; Dridi, Parizeau, Mailhot, &Villeneuve, 2008; El Hayek, Van Voorthuysen, & Kelly, 2005;Fallah-Fini, Rahmandad, Triantis, & de la Garza, 2010; Gharbi &Kenné, 2005; Gomez Urrutia, Hennequin, & Rezg, 2011; Goti,Oyarbide-Zubillaga, & Sánchez, 2007; Guizzi, Gallo, & Zoppoli,2009; Gupta & Lawsirirat, 2006; Hani, Amodeo, Yalaoui, & Chen,2008; Hennequin, Arango, & Rezg, 2009; Ilgin & Tunali, 2007; Johansson & Jägstam, 2010; Kenné & Gharbi, 2004; Kuntz,Christie, & Venkata, 2001; Lei, Liu, Ni, & Lee, 2010; Li & Ni, 2009;Lynch, Adendorff, Yadavalli, & Adetunji, 2013; Marquez, Gupta, &Heguedas, 2003; Murino, Romano, & Zoppoli, 2009; Ng, Lin, &Waller, 2009; Oyarbide-Zubillaga, Goti, & Sánchez, 2007;Oyarbide-Zubillaga, Goti, & Sanchez, 2008; Petrovic´, Šenborn, &Vujošević, 1982; Ramírez-Hernández & Fernandez, 2010;Ramírez-Hernández et al., 2010; Rasmekomen & Parlikad, 2013;Rezg, Xie, & Mati, 2004; Rezg et al., 2005; Richard Cassady,Bowden, Liew, & Pohl, 2000; Roux, Duvivier, Quesnel, & Ramat,2013; Roux, Jamali, Kadi, & Châtelet, 2008; Sarker & Haque,2000; Schutz & Rezg, 2013; Shalaby, Gomaa, & Mohib, 2004;Sheneld, Fleming, Kadirkamanathan, & Allan, 2010; Tateyama,Tateno, & Shimizu, 2010; Triki, Alalawin, & Ghiani, 2013; Um.,Cheon, & Lee, 2011; Van Horenbeek & Pintelon, 2012, 2013;Vardar, Gel, & Fowler, 2007; Wang, Yang, & Schonfeld, 2009;Xiang, Cassady, & Pohl, 2012; Yang, Djurdjanovic, & Ni, 2008; Yaoet al., 2004; Yun, Han, & Park, 2012; Zhou, Djurdjanovic, Ivy, &Ni, 2007; Zhou, Zhang, & Ma, 2012). In order to classify the pub-lished literature and outline main trends in modelling and optimis-ing maintenance systems, each paper was analysed to identifyrelevant features such as application area, maintenance strategiesand policies, simulation modelling techniques and software, opti-misation methods and software, optimisation objectives and deci-sion variables. An abstract version of the analysis for all papers isshown in Appendix A.

3. Overview of reviewed papers

All the papers were published in the year 2000 or after with theexception of one journal paper published in 1982 (Petrović et al.,1982). Fig. 3-1 shows an increasing trend in publications althoughit may not be statistically signicant. These results match thoseobserved in earlier studies, which found that the use of simulationin maintenance is increasing (Alabdulkarim et al., 2013; Garg &Deshmukh, 2006; Sharma et al., 2011). The resulting literaturecomprises of 47 journal articles (80%) and 12 conference papers(20%).

The United States appears to be leading in this research eld fol-lowed by France as illustrated in Fig. 3-2. They both account for

about two-fth of the literature whereas ten countries accountfor the second two-fths.

0

1

2

3

4

5

6

7

8

9

2000 2001 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

n u m

b e r o

f p a p e r s

Year

Papers from 2000 - 2014

Fig. 3-1. Number of publication by year (2000 – 2014) (58 papers).

United States26%

France12%

Belgium7%Canada

6%Spain

5%

Australia4%

China4%

Italy4%

United Kingdom4%

Iran3%

Japan3%

South Korea

3%

Tunisia3%

Other16%

Publica ons by country

Fig. 3-2. Publications by country (59 papers).

0 20 40 60 80 100

Rezg, N.

Allaoui, H.

Ar ba, A.

Xie, X.

Gharbi, A.

Kenne, J.P.

Pohl, E.A.

Cassady, C.R.

Haque, A.

Sarker, R.

Number of cita ons

A u t h o r s

Most cited authors

Fig. 3-3. Most inuential authors.

0

1

2

3

4

5

6

7

Computers andIndustrial

Engineering

Journal of Intelligent

Manufacturing

Journal of Quality in

MaintenanceEngineering

Interna onalJournal of

Produc onEconomics

ComputerAided Civil andInfrastructureEngineering

EuropeanJournal of

Opera onalResearch

Top publica on sources

Fig. 3-4. Top publications sources.

A. Alrabghi, A. Tiwari / Computers & Industrial Engineering 82 (2015) 167–182 169

http://-/?-http://-/?-


4/16

The most inuential authors are shown in Fig. 3-3. Rezg fromLorraine University in France was the most inuential authorpublishing six articles which were cited more than 90 times. Onthe other hand, Allaoui and Artiba from University Lille Nord deFrance published only one article which was cited 88 times. It isinteresting to note that the top four inuential authors work inFrench research groups. In total, around 150 authors contributedto the eld. Around half of them published articles which was citedonly 5 times or less.

The top publications sources are shown in Fig. 3-4. The jour-nal of Computers and Industrial Engineering published morethan any other source. This can be explained by the Industrialengineering nature of the problems in the area, especially theside of simulating manufacturing systems and the applicationsof operation research.

Table 3-1 shows the most eight cited articles. It is interestingto observe that the top ve articles are concerned with jointoptimisation of maintenance and production or spare partsmanagement.

3.1. Application areas

Case studies were conducted in semiconductor manufactur-ing systems (Ramírez-Hernández & Fernandez, 2010; Ramírez-Hernández et al., 2010; Yao et al., 2004), electricity sector(Kuntz et al., 2001; Sheneld et al., 2010) automotive industry(Ilgin & Tunali, 2007; Oyarbide-Zubillaga et al., 2007, 2008),plastic industry (Goti et al., 2007), transportation infrastructure(Cheu et al., 2004; Chootinan et al., 2006; Fallah-Fini et al.,2010; ng et al., 2009; WaNg et al., 2009) and train maintenancefacilities (Hani et al., 2008; Um et al., 2011). It is howeverimportant to note that most researchers tended to use academiccase studies. See for example: (Allaoui & Artiba, 2004; Boschianet al., 2009; Guizzi et al., 2009; Hennequin et al., 2009; Leiet al., 2010; Rezg et al., 2005; Roux et al., 2008; Sarker &Haque, 2000).

While most studies examined maintenance in a production con-text, few researchers examined maintenance operations for work-ing products such as ships or aircrafts. The low number of published papers on military hardware might be due to the poten-tially sensitive nature of these systems. Johansson and Jägstam(2010) suggested an approach to provide decision support formaintenance planning intended for military equipment whileGupta and Lawsirirat (2006) analysed the strategic optimal main-tenance actions for a general multi-component system whosehealth is monitored in real time. Both studies reported the shifttowards product service system as the main motivation for theirresearch. El Hayek et al. (2005) demonstrated the effectiveness of simulation based optimisation for planning maintenanceoperations for an aircraft gas-turbine. It is observed that thereare several differences between maintenance in a production con-text and maintenance in a product service system context. In theformer issues such as bottle necks, buffer size and parts waitingin progress have an impact on maintenance planning. In contrast,logistics and transportation are main issues in product servicesystems.

As observed by Goti et al. (2007) and Van Horenbeek, Buré,Cattrysse, Pintelon, and Vansteenwegen (2013), little research isdirected towards optimising a system composed of severalequipment and most of the research focused on optimising singleequipment without considering the production conguration.Indeed, systems compromising of a single machine producing asingle product (Hennequin et al., 2009) or two exactly identicalmachines (Boulet et al., 2009; Murino et al., 2009) are oversimpli-ed and do not reect the complexity and interactions in real man-ufacturing systems.

3.2. Maintenance strategies and policies

Maintenance strategies can generally be categorised intoCorrective Maintenance (CM), Preventive Maintenance (PM) andCondition Based Maintenance (CBM). As illustrated in Fig. 3-5,CM occurs when the asset breaks down resulting in unexpected

Table 3-1Top eight articles based on citations.

Publication Title Citations

Allaoui and Artiba (2004) Integrating simulation and optimization to schedule a hybrid ow shop with maintenance constraints 88Sarker and Haque (2000) Optimization of maintenance and spare provisioning policy using simulation 62Richard Cassady et al. (2000) Combining preventive maintenance and statistical process control: a preliminary investigation 55Rezg et al. (2004) Joint optimization of preventive maintenance and inventory control in a production line using simulation 47Gharbi and Kenné (2005) Maintenance scheduling and production control of multiple-machine manufacturing systems 46Yao et al. (2004) Optimal preventive maintenance scheduling in semiconductor manufacturing 46Yang et al. (2008) Maintenance scheduling in manufacturing systems based on predicted machine degradation 40Ng et al. (2009) Optimal long-term infrastructure maintenance planning accounting for trafc dynamics 37

Fig. 3-5. Overview of maintenance strategies in the literature.



5/16

shutdown and high maintenance cost. PM is scheduled in order tominimise the impact of a sudden breakdown. PM usually consumesfewer resources compared to CM and can be accommodated in theproduction plans. In fact PM can be as simple as cleaning lters,lubricating and changing oil preventing a failure of a critical a com-ponent that is costly and takes time to be delivered. Because theoperation schedules and environment change dynamically in thereal world, PM can take place unnecessarily. To ensure PM occursonly when needed, CBM was introduced. This can be either in aform of regular inspections to evaluate the assets’ wear or in theform of sensors streaming data to diagnostic software. Thereforemaintenance tasks can be triggered only when the wear reachesa certain level. It is worth mentioning that CBM is sometimesincluded under the branch of PM (Manzini, Regattieri, Pham, &Ferrari, 2010).

The majority of researchers investigated PM as can be seenfrom Fig. 3-6. This includes policies such as time-based (Rouxet al., 2008; Sarker & Haque, 2000) where PM is scheduled everyx units of time or age-based (Chen et al., 2006; Rezg et al., 2005)where PM is scheduled every x units of operating time. Othervariations of preventive maintenance policies include groupblock replacements for unrepairable systems where componentswill be replaced if it fails whereas all other components in thesystem will be replaced at predetermined interval and combinedblock replacements where all components will be replaced atpredetermined intervals but if a component fails, it will bereplaced as well as all components in an operation state (Rouxet al., 2008).

CBM received less attention perhaps because it is relativelynew. However, sensors are becoming lower in terms of cost whichis encouraging the implementation of condition based mainte-nance (Xiang et al., 2012). CBM is becoming increasingly popularespecially in Product Service Systems or long-term services agree-ment where sensors are installed on products to monitor degrada-tion (Gupta & Lawsirirat, 2006). Periodic inspections are analternative to sensors but its frequency has to be optimised as itwill consume resources and affect performance (Kuntz et al.,2001). Van Horenbeek and Pintelon (2012) investigated prognosticmaintenance which is essentially CBM combined with the abilityto predict the deterioration of components in the system to see if it is expected to reach the threshold before the next scheduledinspection; If it does then it is replaced immediately. Although

the applications of CBM are increasing in the industry (Wang,2008), it is evident that it is poorly covered in the literature.

Opportunistic maintenance is a policy relevant particularly insituations where down-time is very costly and a shut-down canbe exploited to perform other maintenance actions. Murino et al.(2009) examined opportunistic maintenance in a continuous pro-duction system where stopping one machine could mean bringingthe whole production system to a halt. Sheneld et al. (2010)examined a eet of aero-engines where unscheduled maintenanceresults in cancelled ights and losing customers.

In reviewing the literature, only limited effort was found to bedirected towards comparing different maintenance strategies andpolicies. Xiang et al. (2012) and Yang et al. (2008) studied a repair-able system where preventive maintenance and condition-basedmaintenance policies were investigated. The focus of Allaoui andArtiba (2004) research was on evaluating the effect of variouspriority rules and heuristics on maintenance scheduling. VanHorenbeek and Pintelon (2012) compared the effect of ve differ-ent maintenance strategies on one machine, namely CM, blockbased PM, age based PM, inspection based CBM and CBM with continuous monitoring.

However, on the whole the research is limited in terms of covering main maintenance decisions such as comparing andselecting the optimum maintenance policies in multi-componentsystems and determining the optimum maintenance resources. Inparticular, investigating the implications of implementing newCBM strategies in manufacturing systems compared with tradi-tional PM policies. In addition, there is a potential of evaluatingheuristics against priority rules set by various optimisationalgorithms.

4. Simulating and modelling maintenance systems

4.1. Modelling maintenance systems

It is interesting to observe that the scope of the maintenancemodels varied signicantly in the literature. The main themesare presented in Fig. 4-1. For instance, Gupta and Lawsirirat(2006) modelled only the assets deterioration, Sarker andHaque (2000) added maintenance resources such as spare partsmanagement and Arab et al. (2013) added production dynamicssuch as buffer capacity. The decision of including an elementshould depend on the level of effect it has on the desired simu-lation output (Robinson, 2007). Although maintenance resources

CM10%

PM70%

CBM20%

Maintenance types in the literature

Fig. 3-6. Maintenance types in the literature.

Assets

Maintenanceresources

Produc ndynamics

Fig. 4-1. Scope of maintenance simulation models in the literature.


http://-/?-


6/16

such as technicians, spare parts and equipment have a directeffect on maintenance cost and scheduling (Bruzzone & Bocca,2012; Duffuaa et al., 2001; Verma, Srividya, & Gaonkar, 2007),only few researchers incorporated them in the simulation model.In fact, the assumption of readily available maintenanceresources is fairly common (Allaoui & Artiba, 2004; Arab et al.,2013; Rezg et al., 2005; Roux et al., 2008, 2013).

Three main levels of modelling assets details are observed inthe literature. The majority of researchers modelled assets as awhole unit. Therefore, the deterioration, failure and interactionon a subsystem or a component level is not modelled in thesimulation. On the other hand, some researchers modelledmachines as subsystems. Oyarbide-Zubillaga et al. (2008) mod-elled assets as subsystems based on types of maintenance activ-ities such as electric/electronic and hydraulic subsystems. Zhouet al. (2012) optimised maintenance for sub-systems connectedin series considering the economic dependency, where carryingmaintenance tasks in groups cost less or more than carrying itindividually. Van Horenbeek & Pintelon, 2012 modelled onlyone subsystem in several machines considering economic, struc-tural and stochastic dependencies. In a more detailed modellingof assets, Roux et al. (2008) evaluated three maintenance poli-cies for a system comprising of two independent components.Sarker and Haque (2000) optimised maintenance and spare partprovisioning policy for 13 identical and independentcomponents.

Gupta and Lawsirirat (2006) highlight the fact that meaning of the term ‘component’ differs depending on the context. It is notpossible to model a complex system comprising of thousands partsfor practical constraints. Therefore it is proposed to consider thecomponents that have signicant impact on the asset performance.Tools such as Failure Modes and Effects Analysis (FMEA) that uti-lise historical maintenance data can be used to identify the mostcritical components.

Modelling identical units while assuming there are nodependencies between them is one of the assumptions researchersconsider to simplify the maintenance system. Other relaxingassumptions include:

Perfect inspections: inspections reveal instantly the real deteri-oration state of the asset.

Perfect maintenance: maintenance job is done perfectly fromthe rst time and there is no chance of misdiagnosis. It is oftenreferred to as ‘machines are as good as new’ after maintenanceactions.

Duration of maintenance actions is constant and sometimes istotally neglected implicating an instantaneous maintenance.

Costs of all maintenance actions are known and constant.Furthermore, cost of CM is always higher than PM.

Maintenance resources such as spare parts, tools and techni-

cians are always available immediately when needed. Failures are detected instantaneously.

Perhaps the most signicant aspect is the modelling of machineaging process. Some researchers simplied it by designing onlytwo states for the machine, either working or broken (Goti et al.,2007). Additionally, the machine is regarded as good as new afterundertaking maintenance tasks. On the other hand, El Hayeket al. (2005) considered an improvement factor that incorporatesimperfect maintenance. Therefore the machine status after main-tenance tasks will not be regarded as good as new, rather it liessomewhere between a broken machine and a new machinedepending on the random improvement factor. Furthermore, theduration between preventive maintenance tasks is reduced as the

machine ages. To schedule PM, Ramírez-Hernández et al. (2010)modelled a PM window constituting of warning date which is

the earliest time a PM can be conducted, due date which is the sug-gested date for PM and late date which is the latest time to conductPM.

Accurate modelling of machine degradation process becomesessential for examining condition-based maintenance where aninspection is conducted periodically to decide which maintenancetasks should be executed (Wang, 2008). Alternatively, sensorscould provide indicators on machines’ health such as vibrationmagnitude and temperature in real time (Xiang et al., 2012). Whenindicators’ reading exceed a specic threshold, a maintenance taskis triggered. Guizzi et al. (2009) simulated condition based mainte-nance via discrete event simulation. In their study, the limitation of discrete event simulation is overcome by triggering special eventsthat increase the machine wear at predetermined intervals.

4.2. Simulation techniques

Discrete Event Simulation (DES) dominates the literature as itwas used alone or combined with other modelling techniques byaround two thirds of researchers (see Fig. 4-2). This should notcome as a surprise since it is the most popular technique in mod-

elling manufacturing systems including production planning,maintenance and inventory management ( Jahangirian, Eldabi,Naseer, Stergioulas, & Young, 2010). DES is the modelling of a sys-tem in which variables’ state change at specic points in time.Thus, the system is modelled by arranging these changes (calledevents) in a chronological order and the system is updated when-ever an event occurs. However, between events, the systemremains unchanged and time is advanced to the next scheduledevent (Banks, 2010).

Most DES studies utilised process-based specialised simulationsoftware that provide graphical user interface such as Arena (Aliet al., 2008; Chen et al., 2006; El Hayek et al., 2005; Guizzi et al.,2009; Hani et al., 2008; Ilgin & Tunali, 2007; Murino et al., 2009)which is offered by Rockwell Automation, Promodel (Arab et al.,2013; Rezg et al., 2004, 2005; Zhou et al., 2007) which is offeredby Promodel Corporation and Witness (Alrabghi et al., 2013;Oyarbide-Zubillaga et al., 2007, 2008) which is an offering of Lan-ner Group. Other DES studies utilised general-purpose programsand languages such as C++ (Cavory et al., 2001; Ng et al., 2009), Java (Boschian et al., 2009), Matlab (Hennequin et al., 2009) andExcel (Li & Ni, 2009). Specialised simulation software provide sev-eral advantages over general-purpose programs such as rapidmodelling, animation, automatically collected performance mea-sures and statistical analysis (Banks, 2010).

Some researchers developed a hybrid model combining DESwith other modelling techniques to gain further advantages.Alrabghi et al. (2013), Xiang et al. (2012) and Gharbi and Kenné(2005) built a discrete event model to represent the general

61%

7%14%

19%

0%

10%

20%

30%

40%

50%

60%

70%

techniquesDES DES with other Other techniques Not disclosed

% o

f p a p e r s

Simula on techniques in the literature

Fig. 4-2. Simulation techniques in the literature (59 papers).



7/16

manufacturing system with the machine degradation processmodelled as a continuous element to reect the fact that machinesage as time passes by.

On the other hand, simulation techniques other than DES werereported in some articles. This includes agent-based simulation(Lynch et al., 2013; Tateyama et al., 2010; Triki et al., 2013) andcontinuous simulation (Fallah-Fini et al., 2010; Gupta &Lawsirirat, 2006).

It is worth mentioning that a considerable number of research-ers did not disclose the simulation technique or the software usedin the research. This surprisingly includes some recent publications(see for example: Johansson & Jägstam, 2010; Yun et al., 2012;Zhou et al., 2012). Therefore it might not be possible for an inde-pendent researcher to replicate the experiments. In contrast toAndijani and Duffuaa (2002) ndings, this study conrms thatneglecting the simulation technique or language is present in theliterature.

5. Optimising maintenance systems

5.1. Optimisation methods

The results obtained from the analysis of optimisation methodsin the literature are shown in Fig. 5-1. Similar to simulation tech-niques, not all researchers disclosed the optimisation methodsthey used (Petrović et al., 1982; Ramírez-Hernández et al., 2010;Sarker & Haque, 2000; Yao et al., 2004; Zhou et al., 2012). Manualoptimisation was reported in several articles where simulationruns are conducted systematically while manually changing vari-ables values in gradual steps, see for example: (Boulet et al.,2009; Kenné & Gharbi, 2004; Lei et al., 2010; Rezg et al., 2005).As can be expected, a serious weakness with this approach is itslimitation in terms of exploring the search space and number of variables.

On the other hand, classical optimisation methods (Rao, 2009)that are analytical and utilise differential calculus to nd the opti-mal point such as scatter search (Chen et al., 2006), Nelder–Meadmethod (Roux et al., 2008, 2013), cyclic coordinate method(Xiang et al., 2012), the modied Powell method (Marquez et al.,2003), Fibonacci algorithms (Asadzadeh & Azadeh, 2014) and sim-ple local search (Gupta & Lawsirirat, 2006; Triki et al., 2013) wereapplied to simple manufacturing systems. One criticism of much of the literature on optimising maintenance by classical methods isthe lack of analysis of the objective function and the solution space.Therefore the justication and proper selection of the optimisationmethod is sometimes absent.

As the complexity of maintenance systems increased (Garg &Deshmukh, 2006; Nicolai & Dekker, 2008), modern optimisation

methods were utilised as they are more capable of dealing withcomplex problems (Deb, 2005; Rao, 2009). Most of these methodsare based on selected behaviours found in nature and sometimesreferred to as non-traditional methods. As shown above in Fig. 5-1, modern optimisation methods were utilised in around half of the papers becoming the most reported optimisation approach.The pie chart below shows the breakdown of modern optimisationmethods in the literature. It is apparent from this pie chart thatonly two modern optimisation methods were applied namelyGenetic Algorithms (GA) and Simulated Annealing (SA). In fact onlyfew articles reported the use of SA. This reects an opportunity toresearch the suitability of other modern optimisation methods tomaintenance problems (see Fig. 5-2).

It can be seen that by far the most reported modern optimisa-tion method is GA. It is based on the process of natural selectionin biology and it has been applied successfully to a wide varietyof practical optimisation problems (Rao, 2009). In addition, it iswell suited for complex simulation based optimisation where thereis no prior knowledge of the response surface typology (Biethahn &Nissen, 1994; Bäck & Schwefel, 1993).

SA comes from the concept of the annealing process in metal-lurgy to harden metals. Metals are melted in high temperature atthe start and then cooled gradually in a controlled environmentto obtain desired attributes. It can be used to solve various typesof problems including continuous, discrete and mixed-integer(Rao, 2009).

Guizzi et al. (2009) and Murino et al. (2009) approach has asignicant advantage. In their study they utilised OptQuest, a spec-ialised optimisation tool that allows the utilisation of multipleoptimisation algorithms including tabu search, scatter search, inte-ger programming, and neural networks. On the other hand, Aliet al. (2008) utilised different optimisation algorithms includedin the Intelligent System for Simulation and Optimisation software(ISSOP) such as component wise enumeration, quasi gradient strat-egy and GA. Yun et al. (2012) conducted a two steps optimisationprocess where both GA and SA are used respectively.

Fig. 5-3 shows how optimisation methods were utilised in dif-ferent maintenance strategies. The use of modern methods andclassical methods is comparable in both CBM and PM strategies.However, manual optimisation was used in less than 10% of CBMsystems compared to around 20% in PM systems. Optimising CMsystems appears to follow a different pattern where modern meth-ods and classical methods were utilised equally.

51%

19%22%

8%

0%

10%

20%

30%

40%

50%

60%

Modern methods Classical methods Manual Not disclosed

% o

f p a p e r s

Op misa on methods in the literature

Fig. 5-1. Optimisation methods in the literature (59 papers).

Gene cAlgorithms

80%

Combina on17%

SimulatedAnnealing

3%

0%

Modern op misa on methods in the litrature

Fig. 5-2. Breakdown of the modern optimisation methods in the literature (33papers).


http://-/?-http://-/?-http://-/?-


8/16

Very limited research was conducted to compare the perfor-mance of multiple optimisation algorithms. Dridi et al. (2008)compared three different variations of GA: Island Genetic Algo-rithm (IGA), Non-dominated Sorting Genetic Algorithm-II(NSGA-II) and Niched Pareto Genetic Algorithm 2 (NPGA-2) ona pipe renewal system. They concluded that the algorithms per-formance varies based on the size of the pipe network. On theother hand, Alrabghi et al. (2013) compared SA with Hill Climband Random Solutions. They found that SA achieved betterresults although its computation time was relatively high.

5.2. Problem formulation for optimisation

An optimisation problem can be described by three main ele-ments: design variables, constraints and objective functions. Eachwill be discussed in details in the following sections,

5.2.1. Optimisation objectivesMinimising cost was reported as an objective in more than 70%

of the studies (see Fig. 5-4). Machines and equipment can be over-maintained which increases preventive maintenance cost orunder-maintained, increasing failure rate and its consequences.Usually reactive maintenance is xed at a higher cost than preven-tive maintenance and the objective is to minimise the total main-tenance cost (Gupta & Lawsirirat, 2006; Roux et al., 2008; Xianget al., 2012). Arab et al. (2013) correctly argues that maintenanceis a part of the manufacturing system and considering mainte-nancecost alone is not sufcient.To counter that, some researchersdeveloped an objective function that encompasses the total systemcost. This might include a penalty for each time unit a machine isunavailable (Alrabghi et al., 2013; Kuntz et al., 2001), the cost of defective products (Oyarbide-Zubillaga et al., 2008), a penalty fornot meeting demand (Kenné & Gharbi, 2004; Rezg et al., 2004) or

spare parts management costs (Ilgin & Tunali, 2007; Sarker &Haque, 2000).

Instead of maintenance cost, Roux et al. (2013) identied max-imising machines availability as the optimisation objective. Theyargue that it is more appropriate as production costs are muchhigher than maintenance costs. Such explanation tends to overlookthe fact that maintenance costs are signicant (Yao et al., 2004)and can be higher than production costs (Wang et al., 2008). Sim-ilarly, Boulet et al. (2009) maximised availability and maintenancecosts were considered manually for each case after the optimisa-tion results.

However, maximum machine availability does not necessarilylead to maximum production throughput in manufacturing set-tings, which is an optimisation objective in several recent studies(Ali et al., 2008; Arab et al., 2013; Lei et al., 2010; Ramírez-Hernández et al., 2010). A machine can be available but not inworking state due to many reasons such as shortage of rawmaterial or blockages as a result of bottle necks. Therefore it issuggested to consider the manufacturing system as a whole andmaximise the production throughput.

In addition to minimising costs, maximising availability andmaximising production throughput, other optimisation objectiveswere identied in the literature. Oyarbide-Zubillaga et al. (2008)considered a more holistic approach where the total cost andprot of the system is evaluated. The costs of maintenance tasksas well as defective products contribute to the cost functionwhereas the prot is calculated by the number of non-defectiveitems produced. The variation in selecting the optimisationobjectives might be due to the nature and purpose of the study.For instance, Ramírez-Hernández and Fernandez (2010) formu-lated the optimisation objective purely on production measuresnamely to minimise both machine cycle time and wait in pro-gress. The purpose of study could have been to support a qualityinitiative without a particular interest in cutting maintenanceresources in the factory. On the contrary, Hani et al. (2008)examined a train maintenance facility where the focus was onminimising the parts immobilization time as well as minimisingoccupation rates for maintenance workshops. Nevertheless,limited discussion of the optimisation objectives choice wasapparent in the literature.

Similar to the situation in many engineering problems(Belegundu & Chandrupatla, 2011), maintenance systems mightrequire optimising several objectives simultaneously such as min-imising maintenance costs and maximising assets availability. It isobserved that researchers used one of the following approaches tosolve that:

Including multiple objectives in one objective function. Forexample, calculating machine downtime as costs (Alrabghiet al., 2013; Kuntz et al., 2001) or including a penalty for

not meeting demand in the cost function (Kenné & Gharbi,2004; Rezg et al., 2004). However, a challenge with thisapproach is transforming an objective in another objective’sunit. For example, estimating how much unavailability of cer-tain equipment would cost or estimating how costly it is tofail to meet the demand. Moreover, these costs are likely tochange depending on the market dynamics (Van Horenbeek& Pintelon, 2012).

Developing a desirability function where optimisation objec-tives are assigned weights according to its importance to thedecision maker to reach the best compromise (Boschian et al.,2009; Boulet et al., 2009; Chootinan et al., 2006; Marquezet al., 2003). This approach does not require transforming anobjective in another objective’s unit. Nevertheless, it forces

the decision maker to trade-off between objectives by assigningweights and ultimately producing a single result.

0%

10%

20%

30%

40%

50%

60%

70%

Modern methods Classical methods Manual

%

o f p a p e r s

Op misa on methods applica ons inmaintenance strategies

CBM

PM

CM

Fig. 5-3. Optimisation methods application in maintenance strategies.

71%

17%12% 12%

0%10%20%30%40%50%60%70%80%

min cost max availability max throughput maxrevenue/prot

% o

f p a p e r s

Most reported op misa onobjec ves

Fig. 5-4. Most reported optimisation objectives (59 papers).



9/16

Utilising multi-objective optimisation algorithms that have theability to solve multiple objectives simultaneously. For instance,Non-dominated Sorting Genetic Algorithm was implemented tominimise costs and maximise prots (Goti et al., 2007;Oyarbide-Zubillaga et al., 2007, 2008). It is interesting to notethat only a limited number of researchers utilised multi-objec-tive optimisation as shown in Fig. 5-5.

5.2.2. Decision variables

Five decision variables were identied as the most reported inthe literature as illustrated in Fig. 5-6. Determining how frequentshould assets be maintained to achieve the best possible solutionis a continuing concern within the eld. It is the most obviousoption in cases where PM is modelled in the system as it can becontrolled and its effect on cost and availability is widely accepted.

However, when the system in interest incorporates CBM (Guizziet al., 2009; Van Horenbeek & Pintelon, 2012; Xiang et al., 2012) oropportunistic maintenance (Sheneld et al., 2010; Zhou et al.,2012), the obvious decision variable becomes the maintenancethreshold that triggers maintenance actions. If information onassets degradation is not streamed by on-line systems, inspectionsare needed to evaluate the degradation of assets. Inspection inter-vals were included as a decision variable in some publications

(Kuntz et al., 2001; Van Horenbeek & Pintelon, 2012; Xiang et al.,2012).

In addition, some researchers optimised maintenance queuingand priority rules for different assets (Hani et al., 2008; Li & Ni,2009). For example, if more than one machine breaks down orrequires preventive maintenance at any given time, which oneshould be maintained rst. It could be that machines in a bottle-neck should have a higher priority to enhance the total throughput.It is another signicant variable that received little attention. Thismay be due to the fact that maintenance resources were not con-sidered in the simulation model so resource usage is not a con-straint. However, it is evident that assigning different prioritiesto machines when maintenance resources are occupied have adirect effect on maintenance performance (Lei et al., 2010;Ramírez-Hernández & Fernandez, 2010).

Spare parts management is an important component in themaintenance system and has a considerable impact on costand availability. Several studies showed that optimising mainte-nance and spare parts policies jointly led to better results com-pared to optimising them separately (Bruzzone & Bocca, 2012;Sarker & Haque, 2000; Zohrul Kabir & Farrash, 1996). Absenceof spare parts when assets are broken extends unavailability.Whereas keeping a large inventory of spare parts results inhigher costs.

Several attempts have been made to investigate the effect of production parameters on maintenance systems in manufactur-ing settings. The work of several authors (Azadivar & Shu,1998; Lei et al., 2010; Rezg et al., 2005) show that buffer sizehave an impact on the performance of maintenance operations.The availability of buffer between machines allows maintenanceresources to be stretched on a longer time with lesser effect onproduction rates. On the other hand, quality initiatives such aslean, six sigma and Just In Time requires the minimisation of Wait-In-Progress.

Researchers have not treated maintenance resources in muchdetail. Only few included maintenance technicians (Alrabghiet al., 2013; Tateyama et al., 2010; Triki et al., 2013; Vardar et al.,2007) or maintenance equipment ( Johansson & Jägstam, 2010) asdecision variables.

5.2.3. ConstraintsConstraints are placed on values a decision variable can take

(Oyarbide-Zubillaga et al., 2008) or the decision variable value inrelation to other variables in the system such as having the maxi-mum stock level of a spare part should be always larger than thereorder point (Chen et al., 2006). Alternatively, constraints can beplaced at other variables such as the maximum budget that canbe spent (Ng et al., 2009), minimum reliability level (Schutz &Rezg, 2013) or PM window where PM actions have to be takenfor each machine (Arab et al., 2013). However, it is common tonot explicitly dene constraints, see for example: (Cheu et al.,2004; Guizzi et al., 2009; Hani et al., 2008).

6. Discussion

Simulation based optimisation has the potential to solve theincreasingly complex and dynamic nature of maintenanceproblems and there is an increasing trend of using simulationto optimise maintenance systems. The current study found thatonly few real life case studies were published, the academiccases that dominate the literature such as a single machine pro-ducing a single product are oversimplied and do not reect thecomplexity and interactions in real systems. Moreover, littleresearch is directed towards optimising a system composed of several equipment and most of the research focused on optimis-

ing few equipments without considering the operationconguration.

Mul -objec veop miza on

15%

Single-objec veop miza on

85%

Single-objec ve vs. mul -objec veop miza on

Fig. 5-5. Single-objective vs. multi-objective optimisation (59 papers).

59%

15% 14%8% 8%

0%

10%

20%

30%

40%

50%

60%

70%

PM frequency Spare parts Maintenancethreshold

Maintenancepriori es

Buffer size

% o

f p a p e r s

Most reported decision variables

Fig. 5-6. Most reported decision variables in the literature (59 papers).


http://-/?-http://-/?-http://-/?-


10/16

A range of simulation based optimisation applications inmaintenance systems across various industries were covered.However, few researchers examined maintenance operations inproduct service systems such as aircraft gas-turbine and militaryequipment.

Very little was found in the literature on comparing and select-ing the optimum maintenance strategy. The majority of research-ers investigated variations of PM including time-based PM andage-based PM. However, investigating condition based mainte-nance as a strategy in a production context is poorly covered inthe literature.

In general, data availability does not seem to be a challenge forresearchers modelling CM and PM systems. Operational data suchas cycle times and arrival patterns for raw material can be obtainedfrom the eld records. Likewise, historical maintenance data suchas breakdown patterns and repair times are commonly available.Cost of maintenance actions are usually simplied by using thecompany’s standards or using calculating the hour rate based onsalaries data.

However, obtaining data on the dependency between compo-nents appears to be a challenge. For example, estimating the effectof the failure of one component on the degradation of connectedcomponents. Gupta and Lawsirirat (2006) suggested a dependencyfactor that is estimated using Failure Mode and Effect Analysis(FMEA), operational data and experts and vendors judgements. Inaddition, a challenge appears when attempting to model themachine degradation in CBM systems. CBM systems based onvisual inspection can be simplied by assuming several xed statesfor the asset where the transition from a state to another is basedon probabilities obtained from the historical records (Kuntz et al.,2001). On the other hand, CBM systems based on on-line sensorsdata are modelled by tting the data into a curve and assumingit correctly reects the change in asset’s health over time (VanHorenbeek & Pintelon, 2012).

Uncertainty is an inherited feature of maintenance systems.Assets’ degradation depends on many factors leading to unex-pected breakdowns. Human errors during inspection or mainte-nance can add signicantly to this uncertainty. Fitting the datainto statistical distributions and then sampling randomly from isa common practise used to account for this uncertainty. Specialuncertainty parameters that account for human error in visualinspection can be introduced. For example, the longer the crackis on a pipe the more likely that it will be detected correctly(Beaurepaire, Valdebenito, Schuëller, & Jensen, 2012). Hennequinet al. (2009) integrated fuzzy logic in the simulation to modelimperfect maintenance actions according to the different skills lev-els of maintenance technicians.

Sensitivity analysis is used to test the robustness of optimisa-tion results in the presence of uncertainty. It helps in evaluating

the optimal solution and make the required modications espe-cially in areas were estimations or simplications have beenmade. For example, investigating how variations in assets’thresholds levels affect the expected cost of the optimal solution(Gupta & Lawsirirat, 2006). Because it is difcult to obtain accu-rate cost data especially for conducting maintenance and inspec-tion activities, it has been subjected to sensitivity analysis inseveral publications (Gomes, Beck, & Haukaas, 2013; Kuntzet al., 2001; Valdebenito & Schuëller, 2010; Zhou et al., 2007).In addition, sensitivity analysis was used to test the robustnessof a suggested model by varying inputs and investigating if theresults are in line with the expected outcome (Boulet et al.,2009).

A vast majority of researchers used discrete event simulationto model maintenance operations. Modelling maintenanceresources received little attention and the majority of research-ers assumed it was readily available. On the other hand, modernoptimisation methods such as GA and SA were the mostreported optimisation methods in the literature. Limited researchwas conducted to compare the performance of multiple optimi-sation algorithms. One criticism of much of the literature onoptimising maintenance using classical methods is the lack of analysis of the objective function and the solution space. There-fore the justication and proper selection of the optimisationmethod is sometimes absent.

Minimising cost was reported as an optimisation objective inaround three quarters of the literature. Moreover, limiteddiscussion of the optimisation objectives choice was apparentin the literature. It is observed that researchers used threeapproaches to solve several objectives simultaneously: includingmultiple objectives in one objective function, developing a desir-ability function and utilising multi-objective optimisation algo-rithms. The latter received little attention despite its ability tosolve multiple objectives simultaneously and provide the deci-sion maker with exibility in the maintenance dynamicenvironment.

Fig. 6-1 presents an overview of optimal problem formulationfor different types of maintenance optimisation problems. Somedecision variables depend on the choice of maintenance strategywhile others can be applied to all maintenance systems. In addi-tion, if the problem includes joint optimisation of maintenanceand spare parts the inventory policy parameters can be opti-mised. That could be either the reorder level and maximumstock level or the reorder level and order quantity. On the otherhand, if the problem includes joint optimisation of maintenanceand production dynamics, buffer size can be considered a deci-sion variable. Optimisation objectives do not seem to be affectedby the type of maintenance system or whether a joint optimisa-tion is present.

Produc on

reorderlevel

reorderlevel

maximumstock level

orderquan ty

objec ves

* Joint op misa on refers to the op misa on of maintenance system and spare parts or produc on

technicians equipmentmaintenance

priori esbuffer size

decisionvariables

PMfrequency

maintenanceschedule

inspec onfrequency

maintenancethreshold

PM CBM

Maintenance strategy

Spare parts

min cost, max availability, max throughput

General maintenanceJoint op misa on*

Fig. 6-1. Optimal problem formulation for different types of maintenance optimisation problems.



11/16


12/16


13/16


14/16

Appendix A (continued )

Publication Simulationtechnique

Simulationsoftware

Optimisationmethod

Optimisationsoftware

Mainte-nancestrategy

Realcase?

Decision variables

PMfrequency

Inspectionfrequency

Maintenancepriorities

Spareparts

Maintenancetechnicians

Buffersize

Mainte-nanceequipment

Mananthr

Rezg et al. (2005) DES Promodel Manual and design of experiments

Statgraphics PM No Yes No No No No Yes No

Gharbi and Kenné(2005)

DES ingeneral butcontinuousfor machineaging andinventory of products

Visual slamlanguage

Manual and design of experiments

Statgraphics PM No Yes No No No No No No

Yao et al. (2004) Mixed-integerprogramming

IBM EasyModeler Not disclosed OSL package PM Yes Yes No No No No No No

Allaoui and Artiba(2004)

DES Resource – action– operation

Simulated annealing Psudo code CM No No No No No No No No

Shalaby et al.(2004)

Not disclosed MATLAB GA NotDisclosed

PM Yes Yes No No No No No No

Kenné and Gharbi(2004)

Markovchains

Not disclosed Manual and design of experiments

Statgraphics CM No No No No No No No No

Rezg et al. (2004) DES Promodel GA Not disclosed PM No Yes No No No No No No Cheu et al. (2004) Trafc

simulatorParamics GA Not disclosed PM No No No No No No No No

Marquez et al.(2003)

Systemdynamics

Not disclosed The modied powellmethod

Direct search PM No Yes No No No No Yes No

Cavory et al.(2001)

DES C++ SIM GA Not disclosed PM Yes Yes No No No No No No

Kuntz et al.(2001)

Markovchains

Not disclosed Manually by runningsimulation scenarios

Not disclosed CBM Yes No Yes No No No No No

Richard Cassadyet al. (2000)

DES Not disclosed EA Not disclosed PM No Yes No No No No No No

Sarker and Haque(2000)

DES SIMSCRIPT II.5 Not disclosed Not disclosed PM No Yes No No Yes No No No

Petrović et al.(1982)

Not disclosed Simne Not disclosed Not disclosed CM No No No No Yes No No No


15/16

References

Alabdulkarim, A. A., Ball, P. D., & Tiwari, A. (2013). Applications of simulation inmaintenance research. World Journal of Modelling and Simulation, 9 (1),14–37.

Ali, A., Chen, X., Yang, Z., Lee, J., & Jun, N. (2008). Optimized maintenance design formanufacturing performance improvement using simulation. In Proceedings of the 2008 winter simulation conference, 7–10 December 2008, Miami, USA (pp.1811–1819).

Allaoui, H., & Artiba, A. (2004). Integrating simulation and optimization to schedulea hybrid ow shop with maintenance constraints. Computers and IndustrialEngineering, 47 (4), 431–450.

Alrabghi, A., & Tiwari, A. (2013). A review of simulation-based optimisation inmaintenance operations. In 2013 UKSim 15th international conference oncomputer modelling and simulation 10–12 April 2013, Cambridge, IEEE,Piscataway, NJ (pp. 353–358).

Alrabghi, A., Tiwari, A., & Alabdulkarim, A. (2013). Simulation based optimization of joint maintenance and inventory for multi-components manufacturingsystems. In Proceedings of the 2013 winter simulation conference, 8–12December 2013, Washington DC, IEEE, Piscataway, NJ (pp. 1109–1119).

Amaran, S., Sahinidis, N. V., Sharda, B., & Bury, S. J. (2014). Simulation optimization:A review of algorithms and applications. 4OR, 12(4), 301–333.

Andijani, A., & Duffuaa, S. (2002). Critical evaluation of simulation studies inmaintenance systems. Production Planning and Control, 13 (4), 336–341.

Arab, A., Ismail, N., & Lee, L. S. (2013). Maintenance scheduling incorporatingdynamics of production system and real-time information from workstations. Journal of Intelligent Manufacturing, 24 (4). 695-695⁄⁄⁄⁄ .

Asadzadeh, S. M., & Azadeh, A. (2014). An integrated systemic model foroptimization of condition-based maintenance with human error. ReliabilityEngineering and System Safety, 124 , 117–131.

Azadivar, F., & Shu, J. V. (1998). Use of simulation in optimization of maintenancepolicies. In Proceedings of the 1998 winter simulation conference, 13–16 December 1998, Washington (Vol. 2, pp. 1061–1067).

Bäck, T., & Schwefel, H. (1993). An overview of evolutionary algorithms forparameter optimization. Evolutionary Computation, 1 (1), 1–23.

Banks, J. (2010). Discrete-event system simulation . Upper Saddle River: Pearson.Beaurepaire, P., Valdebenito, M. A., Schuëller, G. I., & Jensen, H. A. (2012). Reliability-

based optimization of maintenance scheduling of mechanical componentsunder fatigue. Computer Methods in Applied Mechanics and Engineering, 221–222 ,24–40.

Belegundu, A. D., & Chandrupatla, T. R. (2011). Optimization concepts and applicationsin engineering . New York: Cambridge University Press.

Biethahn, J., & Nissen, V. (1994). Combinations of simulation and evolutionaryalgorithms in management science and economics. Annals of OperationsResearch, 52 (4), 181–208.

Boschian, V., Rezg, N., & Chelbi, A. (2009). Contribution of simulation to theoptimization of maintenance strategies for a randomly failing productionsystem. European Journal of Operational Research, 197 (3), 1142–1149.

Boulet, J. F., Gharbi, A., & Kenn, J. P. (2009). Multiobjective optimization in anunreliable failure-prone manufacturing system. Journal of Quality inMaintenance Engineering, 15 (4), 397–411.

Briš, R. (2008). Parallel simulation algorithm for maintenance optimization basedon directed Acyclic Graph. Reliability Engineering and System Safety, 93 (6),874–884.

Bruzzone, A., & Bocca, E. (2012). Innovative solutions based on simulation combinedwith optimization techniques for maintenance service management withincomplex systems. International Journal of Modeling, Simulation, and Scientic Computing, 3 (2), 1–23.

Cavory, G., Dupas, R., & Goncalves, G. (2001). A genetic approach to the schedulingof preventive maintenance tasks on a single product manufacturing productionline. International Journal of Production Economics, 74 (1–3), 135–146.

Chen, M. C., Hsu, C. M., & Chen, S. W. (2006). Optimizing joint maintenance andstock provisioning policy for a multi-echelon spare part logistics network. Journal of the Chinese Institute of Industrial Engineers, 23 (4), 289–302.

Cheu, R. L., Wang, Y., & Fwa, T. F. (2004). Genetic algorithm-simulation methodology

for pavement maintenance scheduling. Computer-Aided Civil and InfrastructureEngineering, 19 (6), 446–455.Chootinan, P., Chen, A., Horrocks, M. R., & Bolling, D. (2006). A multi-year pavement

maintenance program using a stochastic simulation-based genetic algorithmapproach. Transportation Research Part A: Policy and Practice, 40 (9), 725–743.

Deb, K. (2005). Optimization for engineering design: Algorithms and examples . NewDelhi: Prentice-Hall.

Dekker, R. (1996). Applications of maintenance optimization models: A review andanalysis. Reliability Engineering and System Safety, 52 (3), 229–240.

Dridi, L., Parizeau, M., Mailhot, A., & Villeneuve, J. P. (2008). Using evolutionaryoptimization techniques for scheduling water pipe renewal considering a shortplanning horizon. Computer-Aided Civil and Infrastructure Engineering, 23 (8),625–635.

Duffuaa, S., Ben-Daya, M., Al-Sultan, K., & Andijani, A. (2001). A generic conceptualsimulation model for maintenance systems. Journal of Quality in MaintenanceEngineering, 7 (3), 207–219.

El Hayek, M., Van Voorthuysen, E., & Kelly, D. W. (2005). Optimizing life cycle cost of complex machinery with rotable modules using simulation. Journal of Quality inMaintenance Engineering, 11 (4), 333–347.

Fallah-Fini, S., Rahmandad, H., Triantis, K., & de la Garza, J. M. (2010). Optimizihighway maintenance operations: Dynamic considerations. System DynamicsReview, 26 (3), 216–238.

Garg, A., & Deshmukh, S. (2006). Maintenance management: Literature review anddirections. Journal of Quality in Maintenance Engineering, 12 (3), 205–238.

Gharbi, A., & Kenné, J. P. (2005). Maintenance scheduling and production control ofmultiple-machine manufacturing systems. Computers and Industrial Engineering,48 (4), 693–707.

Gomes, W. J. S., Beck, A. T., & Haukaas, T. (2013). Optimal inspection planning foonshore pipelines subject to external corrosion. Reliability Engineering & SystemSafety, 118 , 18–27.

Gomez Urrutia, E. D., Hennequin, S., Rezg, N. (2011). Optimization of a failure –Prone manufacturing system with regular preventive maintenance: An IPAapproach. In 18th IFAC world congress, 28 August–2 September 2011, Milan, Italy(Vol. 18, pp. 10422–10427).

Goti, A., Oyarbide-Zubillaga, A., & Sánchez, A. (2007). Optimizing preventivmaintenance by combining discrete event simulation and genetic algorithms.Hydrocarbon Processing, 86 (10). 115–116+118+120–122.

Guizzi, G., Gallo, M., & Zoppoli, P. (2009). Condition based maintenance: Simulationand optimization. In Proceedings of the 8th WSEAS international conference onsystem science and simulation in engineering, ICOSSSE ‘09, 17–19 Oct 2009, Genova,Italy (pp. 319–325).

Gupta, A., & Lawsirirat, C. (2006). Strategically optimum maintenance of monitoring-enabled multi-component systems using continuous-time jumpdeterioration models. Journal of Quality in Maintenance Engineering, 12 (3),306–329.

Hani, Y., Amodeo, L., Yalaoui, F., & Chen, H. (2008). Simulation based optimization oa train maintenance facility. Journal of Intelligent Manufacturing, 19 (3), 293–300.

Hennequin, S., Arango, G., & Rezg, N. (2009). Optimization of imperfect maintenancebased on fuzzy logic for a single-stage single-product production system. Journal of Quality in Maintenance Engineering, 15 (4), 412–429.

Ilgin, M. A., & Tunali, S. (2007). Joint optimization of spare parts inventory andmaintenance policies using genetic algorithms. International Journal of AdvancedManufacturing Technology, 34 (5–6), 594–604.

Jahangirian, M., Eldabi, T., Naseer, A., Stergioulas, L. K., & Young, T. (201Simulation in manufacturing and business: A review. European Journal of Operational Research, 203 (1), 1–13.

Johansson, E. C., & Jägstam, M. (2010). Maintenance planning using simulation-based optimization. In Spring simulation multiconference 2010, SpringSim’10, 11–15 April 2010, Orlando, society for computer simulation international, San Diego, CA(pp. 1–8).

Kenné, J. P., & Gharbi, A. (2004). Stochastic optimal production control problem withcorrective maintenance. Computers and Industrial Engineering, 46 (4 specialissue), 865–875.

Kuntz, P. A., Christie, R. D., & Venkata, S. S. (2001). A reliability centered optimavisual inspection model for distribution feeders. IEEE Transactions on Power Delivery, 16 (4), 718–723.

Lei, Y., Liu, J., Ni, J., & Lee, J. (2010). Production line simulation using STPN maintenance scheduling. Journal of Intelligent Manufacturing, 21 (2), 213–221.Li, L., & Ni, J. (2009). Short-term decision support system for maintenance task

prioritization. International Journal of Production Economics, 121 (1), 195–202.Lynch, P., Adendorff, K., Yadavalli, V. S. S., & Adetunji, O. (2013). Optimal spares an

preventive maintenance frequencies for constrained industrial systems.Computers and Industrial Engineering, 65 (3), 378–387.

Manzini, R., Regattieri, A., Pham, H., & Ferrari, E. (2010). Maintenance for industrialsystems . London: Springer.

Marquez, A. C., Gupta, J. N. D., & Heguedas, A. S. (2003). Maintenance policies forproduction system with constrained production rate and buffer capacity.International Journal of Production Research, 41 (9), 1909–1926.

Marseguerra, M., Zio, E., & Podollini, L. (2002). Condition-based maintenanceoptimization by means of genetic algorithms and Monte Carlo simulation.Reliability Engineering and System Safety, 77 (2), 151–165.

Murino, T., Romano, E., & Zoppoli, P. (2009). Maintenance policies and buffer sizingAn optimization model. WSEAS Transactions on Business and Economics, 6 (1),21–30.

Ng, M. W., Lin, D. Y., & Waller, S. T. (2009). Optimal long-term infrastructuremaintenance planning accounting for trafc dynamics. Computer-Aided Civil andInfrastructure Engineering, 24 (7), 459–469.

Nicolai, R. P., & Dekker, R. (2008). Optimal maintenance of multi-componentsystems: A review. In K. Kobbacy & D. N. Murthy (Eds.), Complex systemmaintenance handbook (pp. 263–286). London: Springer.

Oyarbide-Zubillaga, A., Goti, A., Sánchez, A. (2007). Determination of the optimalmaintenance frequency for a system composed by N-machines by usingdiscrete event simulation and genetic algorithms. In Proceedings of theEuropean safety and reliability conference 2007, ESREL 2007 – Risk, reliabilityand societal safety, 25–27 June 2007, Stavanger, Norway (Vol. 1, pp. 297–304).

Oyarbide-Zubillaga, A., Goti, A., & Sanchez, A. (2008). Preventive maintenancoptimisation of multi-equipment manufacturing systems by combiningdiscrete event simulation and multi-objective evolutionary algorithms.Production Planning and Control, 19 (4), 342–355.

Petrović, R., Šenborn, A., & Vujoševic´, M. (1982). Spares allocation in the presence of uncertainty. European Journal of Operational Research, 11 (1), 77–81.

Ramírez-Hernández, J. A., & Fernandez, E. (2010). Optimization of preventivemaintenance scheduling in semiconductor manufacturing models using a

simulation-based approximate dynamic programming approach. In Proceedings


http://refhub.elsevier.com/S0360-8352(14)00451-3/h0090http://refhub.elsevier.com/S0360-8352(14)00451-3/h0090http://refhub.elsevier.com/S0360-8352(14)00451-3/h0090http://refhub.elsevier.com/S0360-8352(14)00451-3/h0090http://refhub.elsevier.com/S0360-8352(14)00451-3/h0090http://refhub.elsevier.com/S0360-8352(14)00451-3/h0055http://refhub.elsevier.com/S0360-8352(14)00451-3/h0055http://refhub.elsevier.com/S0360-8352(14)00451-3/h0055http://refhub.elsevier.com/S0360-8352(14)00451-3/h0055http://refhub.elsevier.com/S0360-8352(14)00451-3/h0055http://refhub.elsevier.com/S0360-8352(14)00451-3/h0075http://refhub.elsevier.com/S0360-8352(14)00451-3/h0075http://refhub.elsevier.com/S0360-8352(14)00451-3/h0075http://refhub.elsevier.com/S0360-8352(14)00451-3/h0075http://refhub.elsevier.com/S0360-8352(14)00451-3/h0085http://refhub.elsevier.com/S0360-8352(14)00451-3/h0085http://refhub.elsevier.com/S0360-8352(14)00451-3/h0085http://refhub.elsevier.com/S0360-8352(14)00451-3/h0085http://refhub.elsevier.com/S0360-8352(14)00451-3/h0110http://refhub.elsevier.com/S0360-8352(14)00451-3/h0110http://refhub.elsevier.com/S0360-8352(14)00451-3/h0110http://refhub.elsevier.com/S0360-8352(14)00451-3/h0110http://refhub.elsevier.com/S0360-8352(14)00451-3/h0100http://refhub.elsevier.com/S0360-8352(14)00451-3/h0100http://refhub.elsevier.com/S0360-8352(14)00451-3/h0100http://refhub.elsevier.com/S0360-8352(14)00451-3/h0100http://refhub.elsevier.com/S0360-8352(14)00451-3/h0100http://refhub.elsevier.com/S0360-8352(14)00451-3/h0430http://refhub.elsevier.com/S0360-8352(14)00451-3/h0430http://refhub.elsevier.com/S0360-8352(14)00451-3/h0430http://refhub.elsevier.com/S0360-8352(14)00451-3/h0430http://refhub.elsevier.com/S0360-8352(14)00451-3/h0410http://refhub.elsevier.com/S0360-8352(14)00451-3/h0410http://refhub.elsevier.com/S0360-8352(14)00451-3/h0410http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0450http://refhub.elsevier.com/S0360-8352(14)00451-3/h0435http://refhub.elsevier.com/S0360-8352(14)00451-3/h0435http://refhub.elsevier.com/S0360-8352(14)00451-3/h0435http://refhub.elsevier.com/S0360-8352(14)00451-3/h0435http://refhub.elsevier.com/S0360-8352(14)00451-3/h0425http://refhub.elsevier.com/S0360-8352(14)00451-3/h0425http://refhub.elsevier.com/S0360-8352(14)00451-3/h0425http://refhub.elsevier.com/S0360-8352(14)00451-3/h0425http://refhub.elsevier.com/S0360-8352(14)00451-3/h0425http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0215http://refhub.elsevier.com/S0360-8352(14)00451-3/h0210http://refhub.elsevier.com/S0360-8352(14)00451-3/h0210http://refhub.elsevier.com/S0360-8352(14)00451-3/h0210http://refhub.elsevier.com/S0360-8352(14)00451-3/h0210http://refhub.elsevier.com/S0360-8352(14)00451-3/h0210http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0050http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0400http://refhub.elsevier.com/S0360-8352(14)00451-3/h0355http://refhub.elsevier.com/S0360-8352(14)00451-3/h0355http://refhub.elsevier.com/S0360-8352(14)00451-3/h0355http://refhub.elsevier.com/S0360-8352(14)00451-3/h0355http://refhub.elsevier.com/S0360-8352(14)00451-3/h0355http://refhub.elsevier.com/S0360-8352(14)00451-3/h0305http://refhub.elsevier.com/S0360-8352(14)00451-3/h0305http://refhub.elsevier.com/S0360-8352(14)00451-3/h0305http://refhub.elsevier.com/S0360-8352(14)00451-3/h0305http://refhub.elsevier.com/S0360-8352(14)00451-3/h0340http://refhub.elsevier.com/S0360-8352(14)00451-3/h0340http://refhub.elsevier.com/S0360-8352(14)00451-3/h0340http://refhub.elsevier.com/S0360-8352(14)00451-3/h0340http://refhub.elsevier.com/S0360-8352(14)00451-3/h0340http://refhub.elsevier.com/S0360-8352(14)00451-3/h0310http://refhub.elsevier.com/S0360-8352(14)00451-3/h0310http://refhub.elsevier.com/S0360-8352(14)00451-3/h0310http://refhub.elsevier.com/S0360-8352(14)00451-3/h0310http://refhub.elsevier.com/S0360-8352(14)00451-3/h0310http://refhub.elsevier.com/S0360-8352(14)00451-3/h0420http://refhub.elsevier.com/S0360-8352(14)00451-3/h0420http://refhub.elsevier.com/S0360-8352(14)00451-3/h0420http://refhub.elsevier.com/S0360-8352(14)00451-3/h0420http://refhub.elsevier.com/S0360-8352(14)00451-3/h0060http://refhub.elsevier.com/S0360-8352(14)00451-3/h0060http://refhub.elsevier.com/S0360-8352(14)00451-3/h0060http://refhub.elsevier.com/S0360-8352(14)00451-3/h0060http://refhub.elsevier.com/S0360-8352(14)00451-3/h0060http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0265http://refhub.elsevier.com/S0360-8352(14)00451-3/h0030http://refhub.elsevier.com/S0360-8352(14)00451-3/h0030http://refhub.elsevier.com/S0360-8352(14)00451-3/h0030http://refhub.elsevier.com/S0360-8352(14)00451-3/h0030http://refhub.elsevier.com/S0360-8352(14)00451-3/h0030http://refhub.elsevier.com/S0360-8352(14)00451-3/h0315http://refhub.elsevier.com/S0360-8352(14)00451-3/h0315http://refhub.elsevier.com/S0360-8352(14)00451-3/h0315http://refhub.elsevier.com/S0360-8352(14)00451-3/h0315http://refhub.elsevier.com/S0360-8352(14)00451-3/h0315http://refhub.elsevier.com/S0360-8352(14)00451-3/h0205http://refhub.elsevier.com/S0360-8352(14)00451-3/h0205http://refhub.elsevier.com/S0360-8352(14)00451-3/h0205http://refhub.elsevier.com/S0360-8352(14)00451-3/h0205http://refhub.elsevier.com/S0360-8352(14)00451-3/h0205http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0370http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0260http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0020http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0220http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0230http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0045http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0345http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0380http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0140http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0235http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0200http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0350http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0475http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0405http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0290http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0240http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0250http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0040http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0470http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0455http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0320http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(14)00451-3/h0070http://refhub.elsevier.com/S0360-8352(

state of the art in simulation-based optimisation for maintenance systems

Documents