benifits and cost

8/2/2019 Benifits and Cost

1/19

Available online at www.sciencedirect.com

Omega 32 (2004) 5775www.elsevier.com/locate/dsw

The control of the setting up of a predictive maintenance programme using a system of indicators

M. Carmen Carnero Moya

Technical School of Industrial Engineering, University of Castilla-La Mancha, Avda. Camilo Jos e Cela s/n, 13071 Ciudad Real, Spain

Received 14 December 2001; accepted 15 September 2003

Abstract

Predictive maintenance is one of the maintenance policies which is revolutionising the industry, due to the increase insecurity, quality and availability which is on o er to an industrial plant. However, the implantation of a predictive maintenance programme (PMP) is a strategic decision, and to date the analysis and study of questions relative to its setting up, managementand supervision have not been carried out su ciently. This paper proposes a system composed of indicators to control thesetting up of PMPs which should facilitate the early detection of anomalies which can appear during setting up, thus avoidingthe failure of these programmes. The system developed can be considered a predictive control of the PMP.? 2003 Elsevier Ltd. All rights reserved.

Keywords: Predictive maintenance; Control; Indicators; Decision making

1. Introduction

The increase in competition, globalisation of businesses,moves towards total quality management, constant techno-logical changes, the supremacy of security and the implica-tion of industry in environmental questions, are some of thefactors which have brought about great changes in the struc-ture of companies. These modications have been carriedover to the production area, as this is the one most directlyinvolved in the e ciency and sustainability of the indus-trial processes. This concern has been transferred to main-tenance, traditionally considered a source of costs, and now

associated with more strategic issues, from an approxima-tion based on the concept of sustainability, thereby pursuingthe strategic consensus dened in Boyer et al. [ 1].

The implications in production and maintenance [ 2] sug-gest the need to change the focus of maintenance policies,traditionally centred on short term issues (use of resources,costs, etc.) towards the consideration of longer term goals(competitivity, sustainability and strategy).

Corresponding author. Tel.: +34-926-295300; fax: +34-926-295361.

E-mail address: [email protected]

(M.C. Carnero Moya).

Predictive maintenance is a maintenance policy in whichselected physical parameters associated with an operatingmachine are sensored, measured and recorded intermittentlyor continuously for the purpose of reducing, analysing, com- paring and displaying the data and information so obtainedfor support decisions related to the operation and mainte-nance of the machine [ 3].

The benets which can be obtained by introducing a pre-dictive maintenance programme (PMP) are: an increase inthe availability and safety of the plant, improvements in thequality of products [ 4] and of maintenance [ 5] as well as inthe quantity and quality of the information available aboutindustrial machinery, the increase in the programming capacity of maintenance activities, optimisation in manage-ment of the store for spare parts, support in the designand improvement of industrial machinery, [ 6], reduction of maintenance costs and capacity to research the root causesof breakdowns [ 7], improved image as the time needed pre-ceding delivery to the client is reduced, etc. As explainedin Christer et al. [ 8], it is necessary to identify maintenanceneeds in advance in order to maintain the normal function-ing of production systems.

Multiple models have been developed for the optimisa-tion of maintenance. Cassidy et al. [ 9] propose a system

whereby the decision-making centre can choose between

0305-0483/$ - see front matter ? 2003 Elsevier Ltd. All rights reserved.doi:10.1016/j.omega.2003.09.009
mailto:[email protected]:[email protected]


2/19

58 M.C. Carnero Moya/ Omega 32 (2004) 57 75

multiple maintenance options, such as minimal repair of faulty components, replacement of faulty components and preventive maintenance. Murthy et al. [ 10] suggest a modelto obtain optimum decision making in a maintenance serviceoperation. McKone and Schroeder [ 11], describe the factorswhich contribute to the setting up of total productive mainte-nance programs (TPM), determining which contribute mostto the development of maintenance systems. Deris et al. [ 12]and Wang [ 13] describe the timing of programmed replace-ment of components or consumables, and Sheus [ 14] objec-tive is similar in relation to minimum costs. Triantaphyllouet al. [15] develop a model for classifying di erent crite-ria relating to industrial maintenance, including availability,reliability, etc.

Developments have also arisen concerning the decision tocarry out the maintenance process in an individualised way,or co-ordinating maintenance activities in various compo-nents with the aim of minimising maintenance set-up costs,

like the heuristic algorithm presented in Wijnmalen andHontelez [ 16], in Wildeman et al. [ 17] and in Dijkhuizenand Harten [ 18]. However, whilst an e ort has been madeto construct mathematical models for optimising the preven-tive maintenance policy, we should point out the absenceof tools for optimising and checking the predictive mainte-nance policy.

The aim of this paper is to contribute to research by de-veloping a model which will allow to evaluate the setting upof a PMP, an aspect that has not been analysed until now, asresearch in predictive maintenance has focused mainly onthe development of new diagnostic techniques. Hence thereis a marked lack of models for analysing the problems of PMPs with respect to evaluation, management and control.The model proposed will be applied in industrial mainte-nance, an area in which mathematical developments are of complex practical application [ 19] due to the lack of infor-mation, insu cient understanding of mathematical models by the sta required to apply them in an industrial plant, orto the di culty of applying new models when the companydoes not provide additional resources to adapt to the newsituation.

This paper continues in Section 2 with an explanationof some concepts relating to problems in setting up PMPs.Section 3 explains the characteristics of PMPs. Section

4 includes a description of the indicators which makeup the control system for setting up of PMPs. Section 5 presents the main empirical ndings. Section 6 gives theconclusions.

2. Problems in setting up PMPs

The suggestions put forward in Hipkin and De Cock [ 20]regarding the setting up of TPM and reliability centred main-tenance (RCM) programmes can be extended to the settingup of a PMP. PMPs lack a standardised methodology to

facilitate set-up.

The level of e ectiveness of the PMP during the design, planning and adaptation phases is low, since reliable mea-surements of the condition of the industrial machinery arenot available. However, the company management expectsto obtain favourable results, and it therefore has a negativein uence on the other phases in the set-up of the PMP whichthen contributes to the loss of condence in the programme,with the subsequent reduction in resources allotted to PMPor even its elimination [ 21].

An adaptation period should be established between thetheoretical models and procedures of the PMP, in its techni-cal and management aspects, in order for them to adapt to the peculiarities of the industrial organisation. During this stagethere should be unconditional support for the programmefrom the whole organisation.

The processes of acquiring predictive data may be ob-structed by the productive process. Among the aspects to bestudied are: the absence of information, the fact that infor-

mation may have been obtained under di erent processingconditions, the in uence of traditional policies on predictivecontrol parameters, etc.

Companies may aim to maintain the policies and struc-tures established to date alongside innovative policies. How-ever, the PMP set-up team should encourage the suppressionof the traditional maintenance model of the organisation infavour of the introduction of a dynamic system of mainte-nance policies in direct dependence on the predictive policy.

When di erent predictive techniques are applied, the in-formation for obtaining higher results in the PMP is notintegrated [ 22]. The absence of integration may be caused by the fact that each predictive technique is performed bydi erent sections of the organisation, and therefore by dif-ferent sta members, who do not establish any exchange of information. This may also be due to the di culty of havingcomputer programmes which integrate di erent predictivetechniques.

It is essential to carry out a suitable assessment of thedimensions of the project according to the human resourcesand techniques available to the industrial organisation. Thisfactor generally implies the restriction of research to onefamily of machinery or sector of the industrial plant.

Predictive diagnostic techniques require the use of complex mathematical tools which in turn demand the acquisi-

tion and analysis of historical data during the set up period of a PMP. However, the industrial organisation can act well inadvance of the development of a catastrophic breakdown bymeans of the predictive maintenance plan. Nevertheless, thePMP requires historical information about the break-pointswhich dene the limits of evolution towards a new state of the industrial machinery, that is to say unsatisfactorily andunacceptably severe states.

A PMP may lack information tools or e cient automaticmeans for the acquisition and treatment of the predictiveinformation. The potential of the predictive technology isoptimised by the use of coherent information resources at

the technological level of the technique applied. However,


3/19

M.C. Carnero Moya/ Omega 32 (2004) 57 75 59

inferior information resources may be assigned to the pro-gramme due to a lack of condence in its success or be-cause better resources are assigned to the control of the usualmaintenance policies.

A PMP is considered to be able to eliminate the defects of the organisation or those caused by other maintenance poli-cies. Defects in the maintenance unit may have lead to weak-nesses in training, management, document circuits, proce-dures, etc. so that the probability of defects being transmit-ted to the predictive programme is high.

The relevant predictive information is often not integratedin the computer maintenance management systems (CMMS)and enterprise resource planning (ERP) [ 23]. In some casesthis lack of integration is due to deciencies in the selec-tion processes of the CMMS and ERP, which do not havethe appropriate modules for the treatment of this type of information.

It is di cult to quantify the benets of PMPs. The limi-

tations relating to the evaluation of investments in informa-tion technology [ 24] can be extended to the technological programmes of maintenance: the impossibility of quanti-fying inputs and outputs, the time elapsing between thecosts incurred and the benets gained, redistribution anddissipation of benets gained and lack of management. APMP can provide a lot of benets; but these improvementsare not assessed when the costs involved in a PMP areconsidered.

One of the benets of applying a PMP is a reduction inmaintenance costs. However the cost of the maintenanceunit may be higher if the costs associated with the predictive programme are added to normal maintenance costs, unlessthe activities associated with other maintenance policies of the organisation are modied. Only the appropriate mainte-nance can reduce maintenance cost [ 25].

The simplicity of predictive diagnostic methods facilitatesthe application of the PMP, reduces the complexity of thetraining and information required and the time assigned to it.An example of this concept is the exploitation of trend anal-ysis as opposed to spectrum analysis. However, most simplediagnostic techniques prove to be ine cient as they do not provide any information about the fault which is develop-ing and the remaining time interval before the appearanceof a catastrophic breakdown is very short. In contrast, the

usefulness of the information obtained from complex meth-ods of analysis allows us to obtain a greater quantity andquality of advantages than when working with the simplesttechniques. On the other hand, the technical and human re-sources required for their application are greater in the rstcase. As Gagnon and Sheu [ 26] explain technology acqui-sition is a critical decision.

The evaluation of the cost of modifying the traditionalsystem is complex. A fragmentation of the costs of includingthe technological innovation project has been established,which implies the application of a multi-parametrical pre-dictive maintenance policy in set-up and maintenance costs

of the programme. The application of dynamic maintenance

models implies higher maintenance costs compared to staticmodels with a similar level of set-up costs.

Optimum benets are derived from a PMP when it is applied within the framework of a proactive philosophy in-cluding an orientation towards continual improvement. Thistendency towards optimum performance can be seen whenthe PMP is combined with TPM or RCM.

3. Characteristics of a PMP

Due to the current absence of information about the char-acteristics of PMP set-up, two subjects have been studied inthis section: structuring the PMP set up in phases and cate-gorising the technological levels of the PMP.

The PMP set-up is compound by the following phases:

Design and planning phase . This refers to the interval of time from the rst suggestions of modifying maintenance policies by introducing a PMP until the human and tech-nical resources are available to begin the set-up.

Adaptation phase . This is associated with obtaining re-liable measurements of the state of the machinery undercontrol. This period begins 8 months from the start of set-up and is estimated to last 14 months.

Extension or globalisation phase . Once the time neededto get a return on the investment is reached, the number of machines under control is increased or else new objectivesare set. This phase is usually reached after a period of three years from the start of set-up.

Integration phase . With the other maintenance systems

(CMMS, corrective, preventive, etc.) and productionsystems. This aspect is essential, due to the necessaryinter-dependence of all the functions of the plant.

PMPs have been categorised at di erent technologicallevels depending on cost and diagnostic capacity [ 23]:

Level 0 . Set-up carried out using the control of sensitivevariables. The cost is practically zero and the diagnosticcapacity is very low.

Level 1 . Involves the use of elementary instrumentation,like vibrometers or devices to do the crackle test.

Level 2 . Uses more sophisticated instrumentation, likevibration analysers, data processing software orviscometers.

Level 3 . The cost is high as sophisticated analysis ma-chines are in use; diagnostic capacity is excellent.

Since the setting up of a PMP is a complex project whichrequires the participation of the whole organisation in orderto obtain positive results, once a PMP has been set-up, dif-ferent factors must be controlled so that the appearance anddevelopment of deciencies in the PMP can be detected dur-ing set-up, maintenance and extension of the programme.The detection and diagnosis of these irregularities is the

reason for introducing a control system in the PMP. In


4/19


Maintenance Departments the practice of employing a setof indicators has been commonly adopted. This practice isorientated towards traditional maintenance policies, but hasnot as yet been developed for Predictive Maintenance.

4. Control system for set up of a PMPThe control system proposed has been designed for the

control of PMPs in which the techniques of vibration andlubricant analysis are applied, as they are the most exten-sively applied techniques in industrial plants. Nevertheless,the proposed indicators can be applied to PMPs by other predictive techniques like termography, ultrasounds, etc.

It has been necessary to establish normalised conditionsfor the set up of a PMP that allow comparative results to be exchanged between companies. These conditions arerelated to:

1. The proportion of application of the di erent mainte-nance policies which re ects the state of maintenance inthe industrial plant. The approximation of the percentageof activities in each maintenance policy to the percentagethat is considered optimum before the set up of a PMPis a relevant aspect. For the successful development of aPMP, the application of preventive maintenance is con-sidered essential and also that it should have reached astable state, which is dened as one in which the yieldsof the preventive policy have stabilised. As regards the period of application of a preventive maintenance pol-icy the following options exist: none, less than 6 months,

between 6 months and 2 years, between 2 and 5 years,superior to 5 years. None indicates the breach of a re-striction before the set up of a PMP. The values less than6 months and 6 months to 2 years indicate the unstablestate of the preventive policy applied, since it is still inthe phase of development. The optimization of mainte-nance must incorporate the possibility of perfecting aspects of the preventive policy before approaching the setup of a PMP. The values between 2 and 5 years or supe-rior to 5 years re ect the presence of a stable state in the preventive policy. Thus, they are adequate for the set-upof a PMP.

2. The necessity for setting up a PMP depends upon the presence of critical machinery associated with safety,quality, availability or maintenance cost. The amount of critical machinery is a relevant factor that can precipi-tate the decision towards outsourcing, if that quantity isnot close to certain values. With regard to the quantityof critical machinery in the industrial plant, three valueshave been established for the di erent sizes of company(Table 1).

The indicators proposed for the control system of PMPshave been classied in four categories: economic evalua-tion, external and internal quality of the PMP, organisational

structure and evolution through time.

Table 1Quantity of critical machinery in the set up process

Small company Medium size company Large company

Inferior to 10 Inferior to 15 Inferior to 20Between 10 and 20 Between 15 and 40 Between 20 and 50

Superior to 20 Superior to 40 Superior to 100

4.1. Denition of economic evaluation indicators

The factors of economic evaluation are classied into:those associated with the costs generated by the set-up of thePMP and those which quantify the savings resulting fromthe Programme.

The following factors are incorporated within the groupof costs:

Operational costs of the PMP . This includes costs relat-ing to personnel and factors connected with the predictive programme. This indicator includes costs:(a) Due to predictive set-up planning. The engineer in

charge must previously quantify the expected protsof set-up in each of the objectives pursued.

(b) Associated with initial training. This includes the ac-quisition of basic knowledge of predictive techniquesand how to use the instrumentation.

(c) For internal personnel working on the acquisition andtreatment of data.

(d) Associated with detection and diagnostics of machinery.

(e) For external consultants, when the PMP is not carriedout with the companys own resources.

(f) For continuous training.(g) Due to feed-back of the predictive set-up for example

by training improvement groups. The set-up should be taken advantage for start-up but also as a trial base for improving basic planning. Thus the negativeimplications of PMP design are minimised in earlier phases of the set-up so that the direct repercussion incosts is much lower. These costs should be reducedas the set-up progresses due to the obvious reductionin problems and a continuing trend towards optimum

performance. Instrumentation costs . These are dened as the costs of

redeeming the investment in instrumentation. This in-cludes costs corresponding to:(a) Vibrometers, data collectors and spectral

analysers.(b) Instrumentation for the analysis of lubricants like:

devices to measure the water content or KarlFischer, particle counters, iron particle separators,viscometers, infrared spectrometers, etc.

(c) Software programmes. These are predictive data pro-cessing programmes and can provide diagnoses about

the state of the machinery.


5/19


(d) Adaptation of the computer hardware to the needs of the PMP which has been set up.

(e) Sensors, cables, sensor xing systems, stroboscopiclamps, reactors, lters etc.

(f) Updating of instrumentation and software.

Di erent estimations have been made of the previousindicators according to the technological level chosen forthe PMP; in these estimations the need for or absence of in-trinsic safety measures in instrumentation should be takeninto account.

Indicators concerning the type of prots resulting fromthe PMP include:

(a) Savings or prots from production . By successfullydetecting a fault availability of machinery is increased,which leads to an increase in production.

(b) Savings on safety . The number of anomalies in ma-

chinery is diminished as are the consequences for sta and production resources. Another consequence is thatinsurance costs drop with the set-up of the programme.

(c) Optimisation of the store for spare parts . It should takeinto account:1. A reduction in costs associated with spare parts used

during set up of the predictive programme.2. A reduction in storage costs due to a reduction in the

number of consumables.3. Costs avoided due to the accumulation of devices.

(d) Savings due to the elimination of over-maintenance .(e) Reduction in costs associated with faults caused by the

use of other maintenance policies , especially preventivemaintenance.

To control the costs generated by the PMP maximum andminimum control values can be established depending onthe characteristics of the PMP designed. The programmemanager will be able to check whether the cost generated by the PMP exceeds the standard value.

4.2. Denition of variables and quality control indicators

The function quality has been divided into the followingcategories:

External . Associated with the quality of the process and product, that is to say improvements obtained by meansof the PMP.

Internal . Relating to the quality of the procedures fordetection, diagnosis and correction of irregularities in thePMP that has been set up.

Structural . Related to the organisational aspects of thePMP set-up and which may have a direct impact on theresults obtained via the Programme. However, this factorwill be dealt together with the structural indicators, con-

stituting a PMP self-audit.

The indicators associated with external quality are con-cerned with the modication in opportunity costs, dened aslosses caused by the non-fullment of the functions of theindustrial machinery. Despite the fact that the indicators aredescribed as quality indicators, the evaluation was made byinspecting non-quality.

The indicators are dened as follows:

1. Non-quality level of the product (NQ p). It is suggestedthat the evaluation should be made on the basis of the percentage of products which do not meet the specicquality standard ( D p). Since not all the defects are caused by machines or activities which can be imputed to theMaintenance Department, a parameter has been includedcalled the adaptation coe cient, , with values in theinterval [0,1]. Its value can be assessed in every company although an average value of the interval could be considered adequate. 1 This indicator is quantiedaccording to

NQ p = D p: (1)

2. Evaluation of the productive process (NC process ). Thisincludes: Number of micro-shutdowns caused by the process

which do not necessarily a ect the availability function(NM).

Evaluation of unfullled customer deadlines (UC). Time employed by the production operators to give

notice of breakdowns and unproductive time caused by breakdowns (TB).

NC process

= NM

UC

TB = NM

TB i; j

T i

Q j; (2)

with T i the period of time by which a quantity Q j of the product is late in being delivered.

In the indicators relating to internal quality, have of moreadvanced systems and techniques, encourages us to includean estimation of losses caused by the insu cient e ciencyof the PMP.

This section analyses the factors relating to the quality of the PMP, since the aspects which relate to improved qualityin products and processes, deriving from the programme, areevaluated in the external quality indicators. The indicators

proposed are:

1. Predictive ine ciency costs . These are associated witherrors in the predictive diagnosis and therefore lead tocosts due to low quality in the PMP. Once the PMPhas detected a machine requiring maintenance work, itis necessary to produce the corresponding work order(WO) and to undertake preventative and corrective main-tenance tasks. An irregularity in the diagnosis adds costsin man-hours and a loss of condence in the PMP. To

1 Remaining anomalies are considered to arise from the produc-

tion area.


6/19


quantify the indicator the cost of man-hours spent on the procedure to correct the error is used.

2. Predictive e ciency ( E p). Whilst with the previous indi-cator the aim is to quantify costs caused by low qualityin the PMP, with this one the aim is to evaluate the tech-nical quality of the programme. The ability to carry out

quality diagnoses is what di erentiates an excellent PMPfrom another which will be discarded. This indicator in-forms us about the level of training of the engineers andexperts involved in the programme; its evolution throughtime should exhibit the characteristics shown in the experience curve. For each time period it denes how:

Number of positives failures detected (predictive maintenance) Number of total failures

100: (3)

The total quantity of breakdowns can take on high valuesduring some stages, and at other times will seem insigni-cant. For this reason the previous expression has been mod-

ied as follows:

i

Number of positive failures detected (predictive maintenance) Number of total failures

100: (4)

The values accumulated over the time period i, demon-strate the trend of technical e ciency.

This indicator only makes sense if it includes breakdownsin the industrial machinery included in the PMP.

3. Set-up time ratio (STR ). It is necessary to quantify the period of time in which the management hopes to obtain

results from the PMP, and especially to achieve the in-vestment return period. The objective is to estimate the period of time available to support the PMP. The long in-vestment return period that characterises the PMP is con-sidered. Consequently the following values to evaluatethe set-up time are suggested: 2

Inadequate set-up time: 06 months. Approximate set-up time: 612 months. Optimum set-up time: 1224 months. Approximate set-up time: 24 30 months. Inadequate set-up time: superior to 30 months.

This is due to not achieving the state which relates maxi-mum performance of the PMP to set-up time. The inabil-ity to achieve optimum results in the aforementioned ratiocan be evaluated in terms of costs. However, other aspectsof set-up, like the need to carry it out within a time limitor full objectives planned at the start, allow us to deneapproximate values for the ratio, distanced to a di erentdegree from the optimum value (Fig. 1) .

2 These temporary values correspond to the design and adjust-ment phases of the set up of a PMP. The extension phase is notconsidered because it is possible to make successive extensions of

the PMP.

Fig. 1. Ratio of the optimum and approximate values with respectto the performancetime ratio of PMP set-up.

Fig. 2. Evolution of the e ort for optimum improvement.

4. Optimum improvement e ort (OIE ). Following the sug-gestions set out in [ 27], during the phases of design andset-up of the PMP, continual improvement programmesseek to reduce costs and improve methods. If the ef-fort employed is too high, this may lead to general costswhich are higher than the prots achieved. The search foroptimum improvement e ort represents the improvemente ort which leads to minimum costs (Fig. 2).It is suggested that this should be quantied using thevariations in the indicators predictive e ciency ( E p) and predictive audit ( Au), in which the modications in ef-fectiveness of the PMP are calculated from the technical

and organisational perspectives. To complete the evalu-ation, the global costs associated with the improvement processes used are divided into operational costs ( C o)and instrumentation costs ( C i), where C o and C i are eval-uated once the design and planning phase is complete.After this an ABC analysis is carried out to detect whenthe improvements made are in category C and whentherefore, to limit the improvement programme.

E p + AuC o + C i

; (5)

0 6 E p 6 1; 0 6 Au 6 1; 0 6 C o 6 1; 0 6 C i 6 1;

C o + C i = 0 :


7/19


4.3. Structural indicator

The structural indicator is used to control factors relatingto PMP organisation which will help ensure it is carried outappropriately. This indicator is called predictive audit ( Au ).This is broken down into the following issues:

Attitude. Strategy. Organisational structure. Facilities. Human resources management. Asset management systems. Policies and procedures.

In the category attitude the following variables aredened:

(a) Level of collaboration between departments.(b) Attitude of management during set-up and developmentof the PMP.

(c) Degree of communication between the organiser/manager and the production workers.

(d) Evaluation of the PMP by the production workers.(e) Level of interaction between workers and PMP man-

ager, increase in work absences, number of complaints,verbal abuse, arguments, loss of initiative and creativityof the workers.

(f) Degree of condence in the development of the PMP.

With respect to the factors relating to strategy, the fol-lowing variables are considered:

(a) Number of incentives established to ensure objectivesare achieved.

(b) Establishment of criteria to evaluate changes in thedegree of satisfaction of the client with the PMP.

(c) Level of PMP control measures.(d) Programming of set-up activities, human resources

engineers needed, impact analysis and critical factorsof set-up.

(e) Divergence between the PMP budget and the realannual cost.

(f) If the predictive diagnosis is obtained in an appropriateresponse time from the Production and MaintenanceDepartments.

The organisational structure includes:

(a) If the organisational structure of the PMP is in conso-nance with the strategy.

(b) Degree of continuity of the PMP if the diagnostic,acquisition or management personnel disappear.

(c) Whether actions aimed at minimising PMP overtimehave been organised.

(d) The experience level acquired by the sta of the

programme.

Table 2Variables rating in a PMP self-audit

Qualication ( Ri) Level of fullment

0 Null1 Minimum

2 Limited3 Medium4 High5 Maximum

(e) Determination of the supervision level of the predictiveactivities which have been performed.

Facilities deal with:

(a) Assessment of the predictive scoring obtained.(b) Work space specication (appropriate and su cient) to

perform PMP activities.

With regard to human resources management the follow-ing points are included:

(a) Suitability of provision of training courses, data pro-cessing, procedures and diagnosis of needs.

(b) Description of the responsibilities, functions, etc. of each PMP post.

In the assets management systems:

(a) Postponed activities level in the PMP.(b) Predictive software integration level with expert sys-

tems and the CMMS/ERP of the industrial plant.(c) Level at which the predictive information is assigned to

the Inventory and Purchases Department.(d) Integration level of di erent predictive methods when

this situation occurs.

The procedures and their policies include:

(a) Application level of international and national regula-tions or the internal regulations created in the di erent predictive methods.

(b) If the level and the quality of the predictive maintenance procedures are appropriate.

(c) The level of updating of predictive maintenancedocumentation.

(d) Information layout level (on-line) through computersystems.

All the variables are considered to be of the same impor-tance. Therefore they are given a similar weighting. Wheneach variable is assessed, a rating is assigned ( Ri) as seen

in Table 2.


8/19


Each aspect is assessed through

Rm =

ni=1

Ri Rmax

n; (6)

with n, the number of variables to include in this matter. The

previous expression provides a comparison between each of the aspects to be considered since it is standardised. Theglobal value of indicator is calculated by means of

R =7m=1 Rm

7: (7)

This indicator informs about the improvement e ortwhich has to be applied to the PMP and it immediatelyidenties the organisational deciencies of the Programmecomparing the PMP state with the highest and the lowestlevels.

4.4. Denition of indicators of the evolution through time

of the PMP set-up

When there is the possibility of performing a PMP, ashas been previously explained, the attainment of some objectives which are in accordance with the characteristics of the PMP are generally assessed. However, PMP applica-tion does not guarantee that global advantages will be ob-tained, re ecting that the optimum has been obtained, if itdoes not go hand in hand with a restructuring of the appliedMaintenance policies. The search for the optimum rate of each maintenance policy is the future and current aim of maintenance.

Maintenance Policies Temporal Evolution Index(MPTEI) allows visualisation of this trend towards theoptimum. TAC is dened as the monthly time integrallyassigned to corrective maintenance activities. TAP is themonthly time assigned to preventive maintenance activi-ties and TAPR is the monthly time assigned to predictivemaintenance activities.

TAC = N

i=1

T i i = 1 ; 2; : : : ; N ;

TAP = M

j=1

T j j = 1 ; 2; : : : ; M ;

TAPR = L

k =1

T k k = 1 ; 2; : : : ; L ; (8)

N;M and L represent all corrective, preventive and predic-tive maintenance activities respectively; T i , T j and T k for thetime assigned to each corrective, preventive and predictiveactivity. The characteristics of an optimum PMP have beenanalysed by considering that is at an excellent level duringset-up when the following conditions are veried:

TAP 0; t . Elimination of preventive mainte-

nance policy.

TAPR = a; a = constant ; t . As soon as the stageof adjustment has been reached, and reliable informationis obtained from the PMP, the time destined for the pre-dictive activities tends to stabilize, since procedures foracquisition of predictive information and analysis of thisinformation have been established.

TAC = b; b = constant ; t . In [28] it is exposedthat the corrective maintenance activities are at minimuma certain value, called residual corrective, which remainsconstant.

a b . The quantity of predictive activities must be supe-rior to corrective activities, as if this is not the case, thePMP does not direct the maintenance to be carried outdepending on the condition of the machinery.

b 5(MT) =100; t ; MT = N i=1 T i + M j=1 T j +

Lk =1 T k . The residual corrective is approximately a 5%

of the total maintenance activities. TAPR = TAPR 3elog =log 2 , previous to t , with

TAPR 3 the value corresponding to the predictive ac-tivities in the globalisation phase. is the percentageassociated with the experience in data acquisition anddiagnosis in the PMP. In the globalization phase the re-sults of the PMP must stabilize, and therefore, the timedestined to predictive activities must experience a reduc-tion similar to the typical curves of experience of the production activities.

To obtain the trends towards the optimum of a PMP asoftware has been created in Matlab which simulates the possible situations in PMPs.

The PMP set-up period is another variable to investigate.A reduced set-up stage means a corresponding increase of costs but if it is prolonged it is di cult to determine the ad-vantages that this set-up provides; this is due to excessivetime before the investment is redeemed and due to the gen-eral dissatisfaction of the organisation because the pursuedobjectives are not achieved.

T c is dened as the sum of real time associated with criticalactivities that constitute the PMP set-up. T o is the sum of the activities that constitute the optimum time period of the project, that is to say, when the comparison between thecost of human resources during the set-up stage and the prots obtained by the Programme is the optimum. T m isthe implementation time when the company has to completethe PMP set-up by a specic date.

The following are characteristics which may in uence theadequacy of the ideal temporal programming of the PMPset-up:

Saturation of the PMP set-up manager. Lack of preparation of set-up manager. Programming of unrealistic time periods during design

and planning stage. Unavailability of human resources associated with the

PMP.

Ine ciency of the sta involved in the PMP set-up.


9/19


Lack of co-ordination and mutual understanding level between strategic management.

Lack of mutual understanding and communication be-tween the set-up manager and the rest of the sta involvedin the set-up.

Unavailability of computer tools and support managementfor set-up.

Lack of agreement between external consultants or inter-nal sta with previous knowledge about the PMP set-up.

The proposed objective is the search for an algorithm that provides an optimum solution for the planning of resourceswithin the set-up project development of a PMP. The limi-tations and the environment of the problem are as follows:

1. Several activities are developed to nish the develop-ment of a project; these activities have been previously programmed by the project manager establishing the crit-

icity and the quantity or percentage of each activity thathas to be completed in each phase that constitutes thePMP.

2. Each activity needs to be completed using a series of resources. These can be of several types and with anability that changes for each period of time.

3. Cost minimisation resulting from the anticipation or delayof the project activities with regard to the original plannedPMP. The minimisation of the excess work ability is alsoanalysed.

4. The amount of planned resources, specied in the original programme of the project, must not be exceeded in any period.

The objective is to distribute resources of each type in theset-up phases of the PMP with the aim of not exceeding theavailable ability and of minimising the costs attributed to theearly or late fullment of the periods of the project activitiesif the available ability is not covered in each period. Thiscon ict is dened as earliness tardiness production planningand scheduling system (ETPSP).

The aim is to minimise the objective function cost:

min P

= N

i=1

T

k =1

k

t =1

p i(t ) k

t =1

d i(t )

+

+

k

t =1

d i(t ) k

t =1

p i(t )+

(9)

( x) + = max {0; x}

where N is the number of activities of the project, M thenumber of di erent resources needed for project develop-ment, T the periods of time considered in the planning, d i(k )the percentage of the activity which has to be completed inthe period k , c j (k ) the available ability of the resource j inthe period k , wij the unitary requirements of each type of resource j for each activity i, p i(k ) the percentage of com-

pleteness of the activity i in the period k , the penalisation

coe cient for early delivery periods, and the penalisationcoe cient for late delivery periods.

The following limitations are included:

1. In relation to ability.

N

i=1

wij p i(k ) 6 c j (k ) : (10)

2. Each possible solution must verify that the sum for alldemanded amounts periods of an activity must be similarin input data and in the solutions of the initial populationand in successive generations. That is

k

d i(k ) =k

p i(k ) ; i = 1 ; : : : ; N : (11)

If the number of activities or resources is excessive, thesolution to the problem entails di culties. This problemcan be solved using genetic algorithms (GA) since they aresuitable for solving changing matters, with a large quantityof input data, a large number of variables of di erent nature,etc. Thus providing an optimum solution or an appropriatesolution to the problem. The ability to obtain the solutionto the problem in a reasonable interval of time and with theminimum number of operations has been considered too.

The characteristics of the application of the tool are thefollowing:

Each possible solution is represented by a dimension vec-tor NT; it is a vector whose components have decimalnotation. The elements are the result of the percentage of each activity completed in each time period.

Firstly a group of initial solutions, which are feasiblewithin the limitations previously mentioned, is createdrandomly.

It is necessary to implement the problem cost function insuch a way that the cost value for each solution obtainedcan be achieved, with the aim of evaluating which of thesolutions is best placed to solve the problem.

The estimation of objective function becomes a health orsolution adequacy value. Therefore the healthiest solu-tions will be those closest to the optimum ones. The lowervalues in cost should correspond to the higher values inhealth. To do this, we subtract from the most signicantcost value of all the population the cost of the rest of themembers of the population to obtain the relevant healthvalues for the optimum individuals.

The genetic operators used are: selection, crossing andmutation. These have special features associated with thenature of the problem.

In this case, a xed number of generations introduced as

an input parameter to GA are created.


10/19


After analysing the behaviour of di erent operators, 3 thefollowing are used in the creation of the algorithm accordingto the suitability of the environment and limitations:

Selection with elitism . The mechanism of roulette-wheelselection is adopted for individuals of the population who

have not been regenerated by crossing. To assure thatthe strongest individual stays on, a position in the newgeneration is kept. This acts as a system for obtaining anelite.

Crossing by element . For each pair of individuals whoare crossed for each product or activity, the values of the odd period positions are exchanged, so that additionalsolutions are obtained. In this case, in order to maintain aconstant demand for each product, the di erences betweenthe old values of a solution and the recent ones are kept ina variable; in position T the exchange is not performed, but the di erences variable value is added and the data can be compensated so that the nal demand is kept constant.The ability limitation is also veried.

Simple mutation . For each product of each chromosome,an random number between 0 and 1 is created. If thisnumber is smaller than the crossing percentage, the de-signed mutation mechanism starts up. This mechanismconsists of generating another random number between0 and 1 to select the position ( p ) which is going to bemutated, changing its value by the content in the positionT -p and vice versa (in such a way that it ensures that wecontinue to full the restriction about demands). Then itis necessary to verify, as in the case of crossing that thenew individual created is also feasible.

To complete GA denition, the parameters to be adoptedin the performance must be specied: population size,number of generations, crossing percentage and mutation percentage.

The test-error method is used to determine the quantify-ing of parameters. This technique is appropriate in the caseof crossing percentage and mutation percentage in whichthe adequacy in the inclusion of a higher or lower crossingnumber and mutations can not be determined to nd an op-timum solution. On the other hand in the case of populationsize and the number of generations until the convergence itseems logical to think that the higher these values are thegreater the probability that the obtained result will be closeto the optimum value.

Taking into account the previous considerations and per-formed tests, the conclusions obtained about GA parametersvalues are the following:

Population size : 20. If the number of population individ-uals increases slightly, it strongly intensies the perfor-mance time towards levels of over 60 min.

3 Two-point cross over, single point crossover, mutation in each

element.

Generations number : 50. If this value is amplied the performance time only increases linearly but if thereare many people in the population, this increase is too pronounced.

Crossing percentage : 0.8. The testing analysis with di er-ent percentages of crossing consolidates adjacent valuesto 0.8 are those which provide outstanding results.

Mutation percentage : 0.12. The tests performed showvery similar results in the interval [0.05, 0.15].

5. Results

Next, the results of indicator MPTEI obtained by means of simulation are shown. The results are presented in matrixesthat are suggested as patterns of the control system. The programme manager can compare in each stage the resultsobtained in this programme with the proposed models and

can assess the state of the activities.The matrixes have been created by dividing the percent-

age of activities for each maintenance policy into threecategories. These categories are:

Minimum: 020%. Middle: 20 60%. Outstanding: 60 100%.

This division allows us to organise in columns(denoted bym) the e ciency of the PMP. Value 1 indicates a minimumlevelof predictive activities to attain the optimum state of thePMP. Value 3 indicates a high value of predictive activitiesto attain the optimum condition of the PMP.

In the rows (denoted by l) the output state of each company is explained. 1 indicates lower output conditions of thePMP; 2 species medium layout (50% of corrective main-tenance activities and the remaining 50% of preventive ac-tivities) and 3 indicates an optimum initial situation for thePMP set-up.

In Fig. 3 the evolution of the variables TAC, TAP andTAPR is shown during the phases of a PMP set-up. No ad-ditional resources to those used in the Maintenance Depart-ment before the PMP set-up are considered. In Fig. 3 thefollowing aspects can be appreciated:

Cells 1m . The correctives activities experience a consider-able decline in spite of the fact that does not split the moresuitable previous conditions of set up of a PMP. Residualcorrection is reached when the PMP has stabilised. As a high percentage of predictive activities is needed this may indi-cate that the e ciency of the predictive activities is lower.

Cells 2m . The initial conditions of the PMP show the exis-tence of a similar level of preventive and corrective activi-ties. As in the previous case as high percentage of predictiveactivities is needed the e ciency of the predictive activitiesis lower.

Cells 3m . The initial conditions for the setting up of a PMP

are optimum; that is to say, there is a stable preventive


11/19

Fig. 3. Matrix with the trends towards the optimum of the variables TAC, TAP and TAPR during the stages of a PMP


12/19


maintenance policy and a level of minimum corrective main-tenance. It is possible to eliminate the preventive mainte-nance policy because the maintenance activities stop being performed in relation to temporary considerations, typicalof a preventive policy, and begin to be carried out depend-ing on the state of the machinery. As in the previous casesthe e ciency of the PMP is highest when a low quantity of predictive activities are necessary.

In Fig. 4 the behaviour of each maintenance policy isshown when the resources are intensied in di erent proportions during the PMP set-up. In this gure it is possibleto appreciate in all the cases that an increase in preventiveactivities is produced during the adjustment phase of thePMP. These e ects are considered to be normal since evenreliability has not been reached in the PMP and they lead toan increase in preventive activities. Besides, the e ciencyduring this phase of setting up of the PMP may be low asa consequence of a lack of training and experience of the

personnel of the PMP. Nevertheless, as can be appreciatedwhen the adjustment phase nishes the number of preven-tive activities diminishes, showing that the PMP has reacheda stable phase with reliable predictive data and diagnos-tics. Similar considerations can be appreciated in the cells inFig. 3, although now the e ects of the predictive ine ciencyare greater, due to the increase in resources destined tothe PMP.

Contrary to the previous matrixes, in Fig. 5 the anomaloustendencies during PMP set-up are suggested and it demon-strates how the optimum is not attained. For a better evalua-tion of these situations a PMP with an increased number of resources has been investigated; this circumstance is char-acteristic of PMP management in organisations. With theaim of establishing comparison criteria among situations,the percentage of the predictive maintenance applied activ-ities has been estimated between 10 and 20% as an averagevalue.

In Fig. 5 the di erent anomalous tendencies appreciatedin the setting up of a PMP are presented:

Cell 11 . The setting up of the PMP instead of eliminatingthe preventive maintenance policy based on periodic inspec-tion, acts by increasing the preventive activities. Therefore,the PMP is not applied with the aim of globally improvingthe maintenance area, but as an additional policy, in so do-

ing increasing the costs of the area as it is applied togetherwith the preventive policy.

Cell 12 . The e ect is similar to that of the previous case, but now the predictive activities do not detect failures in themachineries, which means that similar corrective activitiesare maintained. The e ciency of the PMP is very low.

Cell 13 . The setting up of the PMP means that togetherwith the traditional preventive and corrective activities aseries of additional activities are developed in both policiessuggested by the PMP and which are not necessary.

Cell 21 . The PMP so far has practically no in uence overthe preventive and corrective maintenance applied. The ac-

tions suggested by the PMP are not considered.

Cell 22 . The introduction of the PMP does not diminish thecorrective activities and increases the preventive activities.

Cell 23 . The corrective and preventive activities are in-creased by the same amount. The PMP is creating additionalmaintenance activities that are managed together with thetraditional maintenance policies. Therefore, the maintenance policies have not been modied when the PMP has beenset up.

Cell 31 . The set up of a PMP must not increase the correc-tive activities, since the PMP does not have su cient timeto discover the failures from the initial phases of develop-ment; therefore, the detected breakdowns can lead, in mostcases, to the inability to return the machinery to a workingcondition in a short period of time.

Cell 32 . If in the previous case the preventive activitiesexperienced a slight decrease, in this case the preventive policy is not a ected by the PMP, but the corrective ac-tivities experience a relevant increase (greater than in the

previous case). Therefore, the anomalies in the machin-ery have in uenced other components leading to multiple breakdowns.

Cell 33 . In spite of an initial ideal condition for the settingup of a PMP, with a high percentage of preventive activities,the PMP does not eliminate the activities based on time anddoes not reduce the corrective activities. This suggests thatthe increase in preventive activities is causing an increasein corrective activities.

To illustrate the genetic algorithm developed, an example of this application has been created. It starts from the planning of a PMP project according to a Gantt diagram(Fig. 6) of the project, corresponding to a real case. It aimsto distribute the resources in the project so that the costscaused by not meeting deadlines and the costs attributed tonon-covered working ability are minimised. The workingmatrixes are the following:

D(i; k ). Demands matrix for the end of the activities. Itindicates the percentage of the activity i which must becompleted in the period k .

C ( j;k ). Resource ability matrix. It indicates the availableability of the resource j in the period k .

W (i; j ). Resource needs matrix. It indicates the quantityof resources j the activity i needs.

It is considered that nishing a task in advance is not penalised ( = 0) but nishing it after the deadline is.

The cost for not meeting deadlinesand for wasted workingability, in addition to the matrix P (i; k ) representing the percentage of the activity to be completed in the period k arethe result of the algorithm used. In Fig. 6 the convergence of the algorithm can be seen in the optimisation of cost; it stillhas to be ascertained whether the population size and thenumber of generations could be increased to obtain betterresults although the answer to this question would probably be positive. To do this, all that is needed is more performancetime of the algorithm (Fig. 7) .


13/19

Fig. 4. Matrix with the behaviour of each maintenance policy when the resources are intensied in di erent proportions during t


14/19

Fig. 5. Matrix with the anomalous tendencies during the setting up of PMPs.


15/19


Fig. 6. Gantt diagram of the set up of a PMP.

Fig. 7. Genetic algorithm convergence of cost generation in the set-up of a PMP.


16/19


Table 3Results in costs and savings indicators

Company Operational Instrumentation Savings due to the elimination of (activity costs of the PMP costs ($) over-maintenance ($)sector) (annual) ($)

First year of set Second year of setup of a PMP up of a PMP

Petrochemical 198,333 30,050 27,045 29,449Food 1 37,262 0 0 0Pharmaceutical 51,687 0 0 24,040Food 2 49,884 30,050 14,424 14,424Manufacturing of 41,469 0 0 601machinery

Table 4Results in quality indicators and goal reached

Company(activitysector)

Predictiveine ciencycosts ($)

Predictivee ciency

Set-up time ratio The company consideredthat the established objec-tives with the PMP have been reached

Petrochemical 300 83.3% Optimum YesFood 1 601 0.33% Inadequate NoPharmaceutical 150 0.66% Approximate YesFood 2 0 100% Optimum YesManufacturingof machinery

150 0.66% Approximate No

The control system has been checked in di erent compa-nies. A questionnaire was sent to 20 random selected companies. Only ve companies had a PMP. Each of these companies has been identied by means of its sector of activity.The data was gathered using subjectiveobjective questions by means of a 05 scale. The results obtained in some indi-cators are given in Tables 3 5. The lack of data in the companies with a PMP about the setting up process has meantthat some indicators could not be evaluated.

As can be seen in Table 5, the petrochemical plant haseliminated preventive maintenance. The tendency towards

the optimum is due to suitable assignation of resources, con-trol and verication of the results of the PMP and adequatetreatment of the traditional maintenance activities when thePMP is set up. The PMP has been operating for some yearsand the initial investment has been recovered. In the food 1industry there is a low assignation of resources to training,consulting and acquisition of instrumentation that have ledto low predictive e ciency. The in uence of a PMP overcorrective and preventive maintenance is not considered, sothese maintenance policies are maintained and there is a sim-ilar state to that previous to the set up of the PMP. The company does not have a CMMS or historical information about

the machinery and failures. Only 10% of the machinery

of the PMP is critical; consequently, the prots that the PMPcan provide in critical machinery, do not compensate the costthat the PMP generates in non critical machinery. No exten-sions of PMP have been made. In the pharmaceutical company the trend is far from optimum. This is due to the scantymodication of the corrective activities and the permanencyof the preventive maintenance. No reduction in preventiveactivities has been made and this has caused an increase inthe costs associated with the Maintenance Department dueto not having carried out relevant modications in the restof the maintenance policies when setting up the PMP. There

is a delay in obtaining prots in the PMP. In this case thePMP must increase the predictive e ciency in order to ful-ll the expectations of the company as regards the PMP. Todo this it is necessary to have more technical resources. Inthe food 2 industry although preventive maintenance has not been eliminated, it has experienced a notable reduction, ashas corrective maintenance. This shows an evolution of themaintenance policies towards an ideal state. There is high predictive e ciency. This has contributed to a positive eval-uation of the PMP in the company. The preventive policyhas not been eliminated, so it will be necessary to wait oneor two years to discover the nal tendency of the PMP. In

the company dedicated to the manufacturing of machinery


17/19


Table 5Results of MPTEI

Company (activity sector) MPTEI

Petrochemical (Fig. 4; cell31)

Food 1 (Fig. 5; cell21)

Pharmaceutical (Fig. 5; cell11 )

Food 2 (Fig. 3; cell21)

Manufacturing of machinery (Fig. 3; cell12)


18/19


corrective and preventive maintenance have experienced areduction. Nevertheless in spite of a 50% reduction in thecosts of repairing machinery there is a lack of quanticationof the benets regarding safety and quality. This has meantthat the manager does not consider that the aims established, previous to the set-up of the PMP, have been achieved. Nev-ertheless the subcontracting of the PMP in this company mayhave contributed to the PMP being considered to be a tool of preventive and corrective maintenance and not the contrary.

6. Conclusions

The questions related to the structuring of the PMP in phases and the categorising of the technological levels of thePMP have not as yet been analysed despite the importanceof controlling the set-up of a PMP. This paper represents acontribution to this topic.

A control system for the set-up of a PMP composed of in-dicators has been developed. The indicators proposed have been classied in four categories: economic evaluation, ex-ternal and internal quality of the PMP, organisational struc-ture and evolution through time.

These indicators facilitate the early detection of anomalieswhich can appear during set-up, thus avoiding the failure of these programmes. An exploratory study using the controlsystem has been developed in some companies that have setup a PMP. This study demonstrates that a control system isa decision support system in the PMP set-up process.

There are three directions for future research. Firstly,research on the question of developing standards for thesetting-up process of the control system. This contributesto the successful setting up of PMPs. We should also pointout the need to develop standards for the acquisition of theinformation that would allow quantifying of the control in-dicators. Secondly, additional research addressing the ques-tion of how to obtain the levels of control of the di erentindicators in relation to di erent industrial sectors. In so do-ing ideal values can be obtained for each control indicatordepending on the type of company, since the need for aPMP and the available resources are very di erent depend-ing on the type of company and in close relation to the sec-tor of activity. Thirdly, the results of this paper have been

obtained from programs of predictive maintenance based onvibration and lubricant analysis techniques. It would how-ever be very interesting to incorporate new predictive tech-niques such as thermography, ultrasounds, noise and thick-ness analysis. These techniques are generating considerableinterest due to the advantages they o er in thermal, nuclear, petrochemical, electrical plants, etc.

References

[1] Boyer KK, Mcdermott C. Strategic consensus in operationsstrategy. Journal of Operations Management 1999;17(3):

289305.

[2] Jonsson P. Company-wide integration of strategicmaintenance: an empirical analysis. International Journal of Production Economics 1999;6061:15564.

[3] Rao BKN. Handbook of condition monitoring. Oxford:Elsevier; 1996.

[4] Lupinucci M, Perez J, Davila G, Tiseyra L. Improving sheet

metal quality and product throughput with Bentlys MachineryManagement System, Orbit. Bently Nevada 2000;21(3):3741.

[5] Villar JM, Masson LO, Gomes JA. Proactive maintenanceAsuccessful history, Orbit. Bently Nevada 2000;21(3):3341.

[6] Beltran P, Lopez A. El Mantenimento Predictivo en aerogene-radores. Caso practico: estudio de aver as. Proceedings 4

Maintenance Spanish Congress, AEM, Barcelona, 2830 de November, 2000. p. 37785.

[7] Kakkar V. Ontario power generations Nanticoke Power Plant,Orbit. Bently Nevada 1999;20(4):146.

[8] Christer AH, Wang W, Sharp JM. A state space conditionmonitoring model for furnace erosion prediction andreplacement. European Journal of Operational Research

1997;101:114.[9] Cassady RC, Murdock WP, Polh EA. Selective maintenance

for support equipment involving multiple maintenance actions.European Journal of Operational Research 2001;129:2528.

[10] Murthy DNP, Asgharizadeh E. Optimal decision makingin a maintenance service operation. European Journal of Operational Research 1999;116:25973.

[11] Mckone KE, Schroeder RG, Cua KO. Total productivemaintenance: a contextual view. Journal of OperationsManagement 1999;17:12344.

[12] Deris S, Omatu S, Ohta H, Kutar S, Samat PA.Ship maintenance scheduling by genetic algorithm andconstraint-based reasoning. European Journal of OperationalResearch 1999;112:489502.

[13] Wang W. A model of multiple nested inspections at di erentintervals. Computers and Operations Research 2000;27:53958.

[14] Sheu SH. A generalized age and block replacement of a systemsubject to shocks. European Journal of Operational Research1998;108:34562.

[15] Triantaphyllou E, Kovalerchuk B, Mann LJR, Knapp J.Determining the most important criteria in maintenancedecision making. Quality in Maintenance Engineering1997;3:1628.

[16] Wijnmalen DJD, Hontelez JAM. Coordinated condition-basedrepair strategies for components of multi-componentmaintenance system with discounts. European Journal of Operational Research 1997;98:5263.

[17] Wildeman RE, Dekker R, Smit ACJM. A dynamic policyfor grouping maintenance activities. European Journal of Operational Research 1997;99(3):53051.

[18] Dijkhuizen G, Harten A. Optimal clustering of frequency-constrained maintenance jobs with shared set-ups. EuropeanJournal of Operational Research 1997;99(3):55264.

[19] Scarf PA. On the application of mathematical models inmaintenance. European Journal of Operational Research1997;99:493506.

[20] Hipkin IB, De Cock C. TQM and BPR: lessons formaintenance management. Omega 2000;28:27792.

[21] Carnero C, La Torre E, Alcazar MA, Conde J.Control of wear applied to compressor: trends in

lubricant analysis. International Journal on the Science and


19/19


Technology of Friction Lubrication and Wear 1999;225229:90512.

[22] Mobley K. An integrated approach to continuousimprovement. Plant Services Magazine 1997;18(3).

[23] Carnero C. Evaluacion del ciclo de vida de un Programade Mantenimiento Predictivo mediante tecnicas multicriterio.

Thesis, University of Castilla-La Mancha, ETSII, 2001.[24] Shafer SM, Byrd TA. A famework for measuring the e ci-ency of organizational investments in information technologyusing data envelopment analysis. Omega 2000;28:12541.

[25] Hipkin IB, Lockett AG. A study of maintenance technologyimplementation. Omega 1995;23(1):7988.

[26] Gagnon RJ, Sheu Ch. The impact of learning, forgetting andcapacity proles on the acquisition of advanced technology.Omega 2000;28:5176.

[27] Conde J. Trade-o analysis in the workshop logistics.

Proceedings of the Signals and Systems InternationalConference 1989;9:14150.[28] Monchy F. Teor a y practica del mantenimiento industrial.

Masson: Barcelona; 1990.

benifits and cost

Documents