Int. J.

Production Economics 63 (2000) 47}58

Measuring maintenance performance } in search for a maintenance productivity index

Hans Lofsten* K
Department of Industrial Dynamics, Chalmers University of Technology, S-412 96, Goteborg, Sweden ( Received 30 April 1998; accepted 26 October 1998

Abstract A partial maintenance productivity goal is that the "rm should seek to maximize its maintenance productivity in economic terms, and should aim at producing any level of output which is decided upon at minimum maintenance cost with respect to the production system's state. The state of objects or production systems is such a property, i.e., a multidimensional property. In this paper we amalgamate these various dimensions into a single measure of the property involved. All things considered it would seem, that in measuring partial maintenance productivity, minimization of maintenance costs is incorporated as a subgoal, based on the maintenance inputs called for an `optimal budgeta. These imputed maintenance costs do not have to be calculated separately, but emerge as a by-product of "nding a high productivity index. In our partial productivity model, the output prices of the produced products and input prices (maintenance costs) will change over time. Expected changes in the prices of outputs and of current inputs would be built into the model. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Productivity; Preventive maintenance; Corrective maintenance; Availability; State

1. Introduction The purpose of maintenance management is to reduce the adverse e!ects of breakdown and to maximize the production system availability at minimum cost. Production system breakdown causes a loss of output. Traditionally, there are two primary approaches to achieve such a purpose. The objective of preventive maintenance is to reduce the probability of failure in the time period after maintenance has been applied. Another is corrective

* Tel:. #46 31 772 12 30; fax: #46 31 772 12 37; e-mail: halo@mot.chalmers.se.

maintenance, which strives to reduce the severity of equipment failures once they occur. Between these two maintenance tactics, preventive maintenance generally has attracted more attention in industry and academic research [1}5]. Previous research has shown that the use of preventive maintenance procedures may contribute to a rational way of maintaining and operating equipment. Characteristics of the two functions production and maintenance is presented in a simple chart in Table 1 [6]. In addition to these characteristics, both functions must similarly prioritize their activities and provide the right material at the right time. In another study, involving 340 plant engineering and maintenance supervisors, approximately 42%

0925-5273/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 5 - 5 2 7 3 ( 9 8 ) 0 0 2 4 5 - X

48

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

Table 1 A comparison of production and maintenance characteristics [6] Production (1) Many components (2) Many inputs Labour Equipment Overhead Materials (3) Discrete outputs which can be inventoried Maintenance (1) Many components (2) Many inputs Labour Equipment Overhead Materials (3) Diverse outputs which are consumed opon creation

responded that scheduling maintenance activities is one of their three biggest problems [7]. Historically, the relationship between maintenance and production has been characterized by con#ict [8]. Typically, production departments do not readily release equipment for scheduled maintenance. Hence, the "rms maintenance policy may in certain cases consist of only emergency maintenance. Although a "rm may utilize a maintenance policy incorporating scheduled maintenance activities, emergency maintenance cannot be entirely eliminated. In order to determine the levels of maintenance a multitude of information is required. The costs of maintenance, estimated to be between 15% and 40% of production costs [9], and the trend toward automation has forced managers to pay more attention to maintenance. Managers have "gured out, that maintenance, with its high cost and low e$ciency, is one of the last cost saving frontiers in management [4]. This paper provides a conceptual framework as well as some results regarding maintenance policies and development of a maintenance productivity index.

2. Literature review 2.1. Long-run maintenance and investment planning models For one area of maintenance, mathematical modelling and optimization, the literature is extensive. The management science and operations

research journals are replete with optimization models based on statistical analysis of component failures. Traditionally, there are also many corrective maintenance policies to reduce the severity of machine failures [1,10]. Research has examined applications of these corrective maintenance policies. However, studies have primarily focused on the e!ectiveness of one single maintenance policy. The research problem of maintaining industrial production systems has traditionally been modelled in the management science literature as a constrained optimization problem [2,4,11}18]. The purpose of the use of any quantitative discipline is to assist management in decision making by using known facts more e!ectively and by reducing the reliance on subjective judgement. But, in the context of maintenance decision making there is often little factual knowledge available. One of the "rst steps is to determine the objective and once the objective is determined, an evaluative mathematical model can be constructed which enables management to determine the best way to operate the system to achieve the required objective. Two central studies on investments and maintenance are those of Boiteux [12] and of Masse [16]. H The problem of determining both partial replacement or repair and the time for retirement has been dealt by Masse. The problem of how these two H factors, repair and service life should be optimally determined is called by the French the `Boiteuxproblema [12]. To conclude, it can be seen that maintenance planning has been the subject of a large number of studies and a number of the reports have treated preventive maintenance and investment planning explicitly [17}26]. Some of the studies which have dealt with preventive maintenance speci"cally are characterized by the fact that the models are large and complicated. Any interaction between the analysis and the decision makers is thus limited. In other studies, e.g., the problem of determining the optimal service life of a machine has been discussed since a long time in the literature under the assumption that repair cost is given. The analysis of optimal replacement policy was "rst directed towards determining the service life in such a fashion that unit cost of production was minimized [24]. One researcher modi"ed this view

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

49

by maximizing the present value of the machines' output minus its cost [19]. As indicated above, many control actions are open to the maintenance manager. The e!ect of these actions cannot be viewed only from their e!ect on the maintenance department since the consequences of maintenance actions may seriously a!ect other units of an organization. Maintenance engineers in the industry are expected to `copea with maintenance problems without seeking to operate in an optimal manner. For example, many preventive maintenance schemes are put into operation with only a slight, if any, quantitative approach to the scheme. In this paper, the optimization of maintenance decision making may be de"ned as attempting to resolve the con#icts of a decision situation in such a way that the variables under the control of the decision maker take their best-possible values.

period adjusted to real unit terms to facilitate comparison. The index, expressed in real unit terms, should be free of the distortions in units costs from the impact of in#ation. The calculus for choosing between aggregate and component productivity measures is guidelines as to how measure productivity may be gained from an analysis of why we should wish to measure it [29]. The reasons are fourfold: (i) For strategic purposes, in order to compare the global performance of the "rm with that of its competitors or related "rms. (ii) For tactical purposes, to enable management to control the performance of the "rm via the performance of individual sectors of the "rm, either functional or by product. (iii) For planning purposes, to compare the relative bene"ts accruing from the use of di!erent inputs of varying proportions of the same inputs, and (iv) For internal management purposes, such as collective bargaining with trade unions. Various researchers provide the following synthesis that most organizations and their managers are aware that they need to measure productivity for (i) comparing their own performance with that of their competitors, (ii) knowing the relative performance of their individual departments, (iii) comparing relative bene"ts of various inputs, and (iv) collective bargaining purposes while dealing with trade unions [30]. The prescriptions in the articles favoring the aggregate approach to productivity measurement are not carried out in practice [27]. Some researchers states that businessmen see productivity through a wider lens than is customarily used by economic analysts [31]. On the factory #oor, almost any sort of improvement may be counted as a productivity booster. It is not one, however, unless it increases output by more than the increase in inputs } or reduces the amount of inputs needed to produce the same amount of output. When outputs and inputs are counted in dollar terms, this latter technique becomes manifest as a cost reduction program. In fact, it is by way of cost reduction programs that most corporations tackle productivity [31].

2.2. Productivity and ezciency measures in general In general, two methodologies emerge in the productivity literature; the aggregate and the component approaches to measuring productivity [27]. The "rst approach fails to acknowledge both practice and managerial perspectives and the second approach describes practice and draws implications for the design of productivity systems from these observations. These two perspectives on productivity measures are evident in the following remarks [28]: Component productivity measures are designed to measure the performance of a single activity or a relatively small organisational unit. They assist "rst-line-managers in improving productivity. . . Goals are established for the productive use of resources, and actual performance is compared to the determined objectives. Aggregate productivity measures are designed to evaluate the performance of a large collective body (a plant, division, company, or an industry) over an extended time frame. Generally, an aggregate measure is an index that relates current period performance to performance of a base

50

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

Various researchers underline the importance of having measures that can be interpreted by all means to explain the widespread use of operationsbased, rather than "nance-based, productivity measures [32]. Observations of practice led researchers to develop the following prescriptions for practice that had been independently developed in several of the "rms [32]: (i) measurements should be understandable by all organization members, (ii) measurements should be accepted by the individual involved, (iii) rewards and measurements should be compatible and (iv) measures should be result oriented. Some researchers say that productivity measures should be developed for each of the key resources used in the manufacturing process: materials, people, equipment and facilities [33]. The term `e$ciencya is a very loose one [34]. To an engineer e$ciency may mean the ratio output/input or output/theoretical capacity per cent (thus he speaks of the `e$ciencya of a machine), while the cost accountant uses the ratio standard cost/actual cost, per cent or its inverse to measure the `productive e$ciencya of a "rm, department, process or cost center. The economist, when he refers to the e$ciency of a "rm, generally means one or two ratios. The "rst concerns the "rm's success in producing as large as possible an output from a given set of inputs; this he calls productivity, or technical e.ciency. For this purpose it is usual to include as inputs only resources which have physical properties. The second economic measure recognizes the fact that it is not su$cient for a "rm to solve the technical problem of production; it must also suitably adjust its operations to prevailing market prices } in particular, the relative prices of its inputs. Accordingly, the second measure, also a ratio of outputs to inputs, is in value terms. Sometimes it takes the form of the ratio value of output/value of inputs, i.e. product prices as well as input prices are introduced, but more commonly prices are only applied to inputs, to give a measure of e$ciency in terms of unit costs, which may be seen as the product of the "rm's `technical e$ciencya and `price e$ciencya [34]. Economic theory lays down the e$ciency conditions governing the allocation of existing resources (i.e. resources being used in current production) and

future resources. There are several types of measures of e.ciency, which require the concepts of input and output [35]. Input refers to the resources which are consumed or expended in taking a course of action [36]. Thus input refers to the cost of a course of action. Output may be measured in terms of either resources which result from taking the course of action. The type of measure of e$ciency depends on whether the amounts of input and/or output are speci"ed in the de"nition of the relevant outcome [35,37]. The four possibilities are as follows [35]: (i) Speci,ed (or irrelevant) outputs and inputs. Here the measure of e$ciency is simply the probability that the outcome will occur. (ii) Speci,ed outputs and variable inputs. Here alternative courses of action are evaluated with respect to the inputs required to obtain a speci"c output. (iii) <ariable outputs and speci,ed inputs. Here alternative courses of action with speci"ed "xed inputs are evaluated relative to the amount of output they yield. (iv) <ariable inputs and outputs. Here neither input nor output is speci"ed, and their di!erence ratio is used as a measure of e$ciency. The course of action which yields the greatest di!erence between output and input can be considered to be the most e$cient. One of the problems is that a production process does not maintain a constant output per unit of input. Inputs and outputs vary, and this variation must be taken into account.

3. Empirical study + maintenance measurements, inputs and outputs The "eld work for this project is a study of the maintenance measurements in eight industrial "rms. Information was gathered using personal interviews with top management and internal and external documents. The project consisted of several substudies, altogether covered by interviews, questionnaires and published data. The empirical studies are expected to illuminate, in particular, the

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

51

composite nature of the `solutionsa adopted in di!erent industrial facilities. The individual "rms have di!erent histories and production systems. The "rms chosen included two in the energy sector (Gothenburg Energy and Ringhals Nuclear Power Plant), two "rms in the chemical industry (Borealis Stenungsund Site and Akzo Nobel Surface Chemistry), two "rms in the manufacturing industry (Volvo Cars and SKF Industrial Bearing Division), one "rm in the paper manufacturing industry (STORA Molndal) and K one "rm in the maintenance of the rollingstock market (SJ Group, The Mechanical Engineering

Division). A summary of the major empirical "ndings are listed in Table 2. First, productivity and e$ciency are terms that are diversely interpreted, but in practice, generally refer to the relationship between production outputs and maintenance inputs. Based on our observations in the eight "rms, production output in the must useful de"nition in a maintenance perspective, focuses on availability, but the economic evaluation is missing. On the input side, the observed organizations concentrated on the key input resources that were crucial to achieving output objectives.

Table 2 Summary of the major empirical "ndings in the eight "eld studies Firm Availability/ measurement of plant state 97%/no, but inspections Input/output (maintenance) Measurement of productivity/e$ciency or maintenance concept general pro"t measurements, absorption costing, priority of objects General pro"t measurements reliability centered maintenance (RCM) Key performance index (KPI), total preventive maintenance (TPM), mean time between failures (MTBF) Continuous improvements/ management of total quality (MTQ) `Internal e$ciencya allocation of resources Measures of productivity/cyclic planning Main maintenance goal and con#icting goals

Gothenburg Energy Ringhals

Maintenance cost/kWh

Availability/suboptimization of resources Availability/output and safety

77%/yes

Production cost/kWh

Borealis

94%/no, but inspections

Maintenance cost per unit

Availability/planned stops and availability

Akzo Nobel

98}99%/no, but inspections

Maintenance cost per unit

Availability/no clear maintenance goals

STORA Molndal K SJ Group

90}95%/no, but inspections 90%/no, but inspections

Maintenance cost per unit Maintenance cost per km

Volvo Cars

98}99%/no, but inspections

Maintenance cost per unit

SKF

50%/no, but inspections

Maintenance cost/ availability

Overall equipment e!ectiveness (OEE) total process e$ciency (KLT), RCM Machine e$ciency, Availability/no object di!erent scorecards, TPM, priority rules total quality management (TQM)

Availability/planned stops and availability Productivity and availability/preventive maintenance not value-added OEE/many di!erent goals

52

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

Second, the aggregate approach, focuses on the design of total corporate productivity/e$ciency measures using aggregated "nancial data. Practice seems to use individual or component measures to judge short-run e!ectiveness, e$ciency or competitiveness. Third, managers in the eight "rms feel that it is inappropriate or unnecessary to convert `productivity and e$ciencya measures to "nancial measures because either the `productivitya or `e$ciencya measure is capable of signaling improvement and there is therefore no need to quantify this improvement further, or because use of a "nancial measure might obscure performance on an activity that the system is capable of measuring directly. Fourth, data derived from the "nancial accounting process, such as costs, variances and pro"ts, are not used to control the shortrun performance. Traditional cost accounting measures, such as unit costs, are used to evaluate the productivity/e$ciency of the maintenance departments or in comparing the organization's performance in the whole organization.

Total output Partial factor productivity" . Single input

(1)

4. A partial maintenance productivity index 4.1. Dexning objectives The identi"cation of objectives normally consists of naming the objectives and it is a long way between naming these objectives and obtaining suitable measures for the degree to which they are obtained, for all of them are quantitative in character. In this exploratory study, I can supply only a tentative hypothesis to summarize the principles of developing maintenance productivity measures. However, the costs of maintenance, estimated to be between 15% and 40% of production costs and other di$culties, such as "nding all inputs and outputs, organizations use other measures than total productivity [9]. Productivity measurement attempts to answer the basic question of how much input is required to achieve a particular output. This question may be posed for a single machine, manufacturing cell or an entire economy. A useful measure of productivity is partial productivity. This measures the total output divided by one kind of input.

In practice, the `total outputa is taken as the availability and amount of production and does not include secondary outputs. However, there are dif"culties to value the production of products in di+erent stages in the production process. It is important to be aware that the maintenance productivity index gives only one limited view of the organization. To get a broader view several measures should be taken which relate to di!erent aspects of operations. The main objective function is (see the empirical results) the availability of the production system. `Gooda productivity when planning maintenance is vital, but it is also crucial to de"ne what is meant by maintenance productivity. In this paper the term productivity is closely coupled to the planning of maintenance. Good productivity is reached when the total maintenance costs and down-time costs are reduced to a minimum with a given minimum level for the state of the production system, i.e. the minimum production rate is ful,lled. Expressed in other words, our partial productivity index is the output divided by the sum of variable maintenance costs and downtime costs. In its simplest terms, the measure re#ects how good a plant is at turning maintenance inputs into production output in dollars. Were we to measure only labor productivity, we could simply use labor hours alone. We need to convert hours into dollars so that we may add it to other types of inputs. It is assumed (see Fig. 1) that each production system has a couple of bottleneck machines. The model presented in this paper determines whether to schedule preventive maintenance and the model trades o! the capital costs of preventive maintenance and the sum of corrective maintenance and downtime costs based on the operations state. The production system in the production environment consists of either series systems or parallel systems, that are not necessarily identical (and therefore subject to di+erent output rates). Maintenance and reinvestment programmes a!ect the reliability of the production system. If an operation in the production system is poorly designed, no amount of maintenance will make it

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

53

the costs related to a preventive maintenance intervention and the costs deriving from a break-down are known. The latter always exceed the former; preventive maintenance would otherwise not make sense, since it would certainly be more convenient to break down and repair it afterwards. Another concept, utilization, is the ratio of actual output to designed capacity. The designed capacity of the machine is the maximum output that could, ideally, be achieved in a week. This ignores the time needed for maintenance and setups. The e!ective capacity is the maximum output that could reasonably be expected. This takes into account the time needed for maintenance and setups. The aim of capacity planning is to match available capacity to forecast demand, but is not focused in this paper.
Fig. 1. Allocation of preventive maintenance activities to operation n, and the output.

`work bettera. The overall reliability of the production system depends on both the reliability of each component and on the way they are arranged. On reliability and maintenance issues, the literature proposes a variety of techniques to assess the reliability of complex systems, to highlight an optimal maintenance policy, to establish the frequency of inspections and preventive maintenance, etc. [38]. In particular, a lot of practices have been put forward for short-term production scheduling. In each of them, some speci"c performance measures (machine utilization, tardiness, #owtime, etc.) are optimized by using algorithmic techniques or expert system shells. Availability is the degree to which the operation is ready to work. An operation or machine is not available if it has either failed or is being repaired. There are several di!erent ways of measuring availability depending on how many of the reasons for not operating are included. Lack of availability because of planned maintenance or changeovers for example. When production systems are designed so that they run in a trouble-free manner, and can be easily recti"ed when necessary, they are said to have a high maintainability. This can be quanti"ed by the mean time to repair. The measure of reliability, the mean time before failure, can be combined with the mean time to repair to give the overall measure of an operation's availability. In our study,

4.2. Inputs } changes of maintenance in the production system Under `normala conditions it is too complicated to analyse each separate maintenance object, at the same time as having to take the allocation of the limited maintenance resources and "nancial restrictions into consideration. It can thus be appropriate to at "rst carry out an object analysis of each individual maintenance object, i.e. a bottle neck machine (i, j,2, n), in this case an operation or machine n. In the next stage one can then carry out a resource allocation analysis, i.e. one chooses a combination of measures for each object which uses the available limited resources. The inputs are represented by 2 C (PM , CM , S ) R R R R, Inputs "Min L (1#i)R R (2)

where C is the total cost in period t. It is the sum of R the expected maintenance cost and down-time costs, is the reinvestment occuring in time . Expected machine life (in years), t the period t (in years or months), i the interest rate, PM the preR ventive maintenance in period t (in dollars). The variable PM denotes the period in which a prevenR tive maintenance job starts, CM the corrective R maintenance in period t (in dollars), and S the state of the machine. The state of the machine can by

54

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

preventive maintenance be improved (i.e. 0, 1, 2, 2, S ). It is needed because when preven  tive maintenance is performed, the machine is assumed to be nearly as good as `newa, so that the possibility of a disaster solely depends on the last period in which preventive maintenance was performed, and where the period t"(i, j, 2, n) i.e. CM (CM , CM , 2, CM ), and S (S , S , 2, S ) R G H L R G H L where n is a machine in the production system. Through use, the machine deteriorates and this deterioration may be measured by the measurement state. The basic purpose behind an inspection activity is to determine the state of the machine. When the preventive maintenance should take place ought to be in#uenced by the capital costs of preventive maintenance activities which will be related to the indicators used to describe the state of the machine or operation and the bene"ts of the preventive maintenance. There are three types of measures to maintain the "rm's machines at a speci"c state, they are reinvestments, preventive or corrective maintenance. An analysis based on preventive maintenance can be carried out using a number of alternative equidistant points of time t"1, 2, 2, !1, whereas reinvestment can occur at the point in time , where corresponds to the machines economic life. We can di!erentiate two decision variables; the size (cash #ow) of preventive maintenance (PM) in dollars and the point in time for the reinvestment (a new machine) . Preventive maintenance means that the machine is regularly checked, in accordance with preplanned selective measures. The aim of these measures is to prevent major damage from occurring and thus avoid large expenditure. The state of the machine or operation can by preventive maintenance be improved (i.e. 0, 1, 2, 2, S ). Corrective mainten  ance is activities which are of a corrective day-today nature. The state of the machine by corrective maintenance cannot be improved, but a failure can be avoided in the short run. When the machine has a decreasing performance function, the costs of corrective maintenance will be increasing. Minimum total cost of schedule is the sum of the expected maintenance cost and down-time costs. Costs of state include not only corrective maintenance costs but also the down-time costs:

State costs"Corrective maintenance costs #down-time costs. (3)

With the above information gathered, the model proceeds to evaluate the "nancial performance of di!erent maintenance policies with respect to different states. The machine's state implies the service, quality and availability. The performance of the machine's state is measured as 0 (total disaster) to S (maximum level of state). It is necessary to  connect costs to di+erent state levels. The term state does summarize in one numerical index, mainly for analytical and costing convenience for the di!erent types of service levels. From the `common sensea point of view it is considered to be bene"cial to carry out preventive maintenance in accordance with minimum requirement levels for the state of the production system, S *S , t"1, 2, 2, , R  where S (S , S , 2, S ) (4) R G H L or when programmed strategies for preventive maintenance require work to be carried out. It is, however, very important to point out that di!erent machines exhibit very di!erent changes in state. Determination of preventive maintenance decisions for probabilistically failing machines or operations involves a problem of decision making under one main source of uncertainty, namely; it is impossible to predict with certainty when a failure will occur, or more generally, when the transition from one state of the machine to another will occur. A further source of uncertainty is that it may be impossible to determine the state of the machine, unless a de"nite maintenance action is taken. In the case of a constant state of the machine, it may well be that the appropriate procedure is to allow the machine to break down before performing a replacement and this decision can be made simply by obtaining statistics relevant to the machine and does not involve construction and solution of a model to analyse the problem. Machines can be performing and non-performing following three di!erent classi"cations of failure: (i) when the machine or operation has a monotonically decreasing (linear or non-linear) performance function, (ii) the machine can have discrete conditions of performance or

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

55

nonperformance and (iii) the machine can #uctuate between good performance and non-performance. The analysis is only valid for the "rst type of failure, where assumptions are made about a linear or non-linear curve, which represents performance, and is able to accommodate discrete or continuous improvements of performance. The performance function of the machine must be monotonically decreasing (linear or non-linear), before preventive maintenance is worthwhile. However, in practice, much preventive maintenance is predicated on the increasing hazard rate of parts with age. As mentioned earlier, the objective function models the minimization of total costs, consisting of preventive and corrective maintenance costs and down-time costs. To apply the present value analysis we need to obtain estimates of the components of the cash #ow that each maintenance policy generates. While the values of some parameters such as preventive and corrective maintenance are usually available, others are more di$cult to obtain or even estimate, such as down-time costs. From time to time, major surveys of the machine's state, are performed by statutory requirements. Between these surveys the corrective maintenance of the machine increases due to deterioration of fallible parts. The preventive maintenance activities cost money in terms of materials and wages, and a balance is required between the money spent on preventive maintenance and savings obtained by reducing the corrective maintenance and downtime costs (costs of state).

4.3. Output } availability and value of amount of production To make a useful productivity index for comparing plants or di!erent machines or operations requires solving a couple of measurement problems. First, measured outputs must be comparable across machines or production systems; few "rms produce a completely homogeneous output at all their sites. Second, all of the di!erent inputs (capital, labor, materials: preventive maintenance, corrective maintenance and down-time costs) used at each machine must be aggregated into a single `unit inputa. The third measurement problem arises

because not all plants or production systems operate in the same environment. Some may have higher e$ciencies because they enjoy economies of scale or a healthier orderbook. Di!erences in productivity arising from speci"c local factors are therefore of less interest to management. In order to measure the degree to which the output has been maximized it is necessary to select a period with respect to which the measure is to be made. It is clear that in most cases one cannot consider only the shortrange e!ects of a decision but must also consider the long-range consequences. This is usually taken care of by a discounting procedure. The state of the machine depends on (among other things) the condition of the equipment involved. Equipment which is likely to be usable for only a short time does not yield nearly as much readiness as equipment which can be expected to operate over an extended period of time. Furthermore, readiness depends on the ability of the force involved to maintain the machine in an operating condition (S , i.e. minimum require  ment of state) so that some measure of down-time for repair should be a part of the measure. One could continue to point out shortcomings of that index, but it soon becomes apparent that a full-time research e!ort could be directed at only the question of how to measure this state and costs of state adequately. Some properties cannot be represented by a single measure; two or more are required. Such a requirement can be revealed by careful analysis of the meaning of the concept. The state of objects or production systems is such a property, i.e., a multidimensional property (capacity, availability, etc.). In this paper we amalgamate these various dimensions into a single measure of the property involved (see Fig. 2). Availability is one dimension of the multidimensional measure state and we have to transform the state to availability. We have already considered that the inputs (preventive maintenance, and state costs) and outputs (availability, production volume and value of the products in di!erent stages in the production system) associated with a particular course of action will vary, but not all inputs and/or outputs will occur at the same frequency. Output is here measured in dollars, and investigations have shown

56

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K

Fig. 2. State, availability and preventive maintenance.

that the value of money is usually not a linear function of the amount [39]. In the maintenance case, there are only nonlinear functions between preventive maintenance and state, state and availability and so on. Now we need some criterion for selecting that course of action which has the `besta productivity index. Our possibility is to measure the production value in a period t, i.e., the production value (output) at each machine (i, j, 2, n) compared to the maintenance input at the same machine in the production system. Eq. (5) expresses the output of the maintenance policy (where machine n is one operating in the production system: 2 R (< P A ) R L L L, Output " L (1#i)R R (5)

5. Conclusions Whatever form they take, productivity and e$ciency calculations are made for two purposes; to decide upon the allocation of resources, and to evaluate the performance of a business after resources have been committed (planning and control). In our case, it is possible to plan and allocate the maintenance resources to achieve high partial productivity and control aspects. However, we must realize that this is only a partial productivity index, and the "rm may in fact have multiple objectives. A partial maintenance productivity goal is that the "rm should seek to maximize its maintenance productivity in economic terms, and should aim at producing any level of output which is decided upon at minimum maintenance cost with respect to the machine1s state (i.e. availability). All things considered it would seem, that in measuring partial maintenance productivity, minimization of maintenance costs is incorporated as a subgoal, based on the maintenance inputs called for an `optimal budgeta. This important cost minimization subgoal would be sought in di!erent ways by solving the prior `maintenance cost minimization problema through: (1) the business maintenance cost minimization programme, and (2) the dual of the maintenance allocation which minimizes maintenance cost, represented by the constraint (state) in the model

where R is the revenue in period t, < the production volume, and P the price of products in di!erent stages in the production system where t"(i, j, 2, n) and S "f (i, j,2, n). Ideally, one would expand R Eq. (5) by expressing each of the above outcomes as functions of the controllable and uncontrollable variables. Then, one would seek to minimize the total expected maintenance costs and downtime costs with respect to a certain state (and hence availability) of the production system by selecting the proper values of the controllable variables. Table 3 illustrates in a productivity matrix the output n and input n at a machine n.

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K Table 3 An illustrative productivity matrix at a bottleneck machine in the production system Input L Output L Controllable variable availability L Input1 Input2 Input3 Input4 . . . 0.2 0.4 0.6 0.8 Uncontrollable variables Volume L Price L Production value (A<P) L

57

presented. These imputed maintenance costs do not have to be calculated separately, but emerge as a by-product of "nding a high productivity index. It would be very useful in evaluating business decisions if a de"nition of ex ante and ex post could be developed that would also meet the accountants crucial requirement of objectivity. There are also two cases to be considered. In decision problems we are concerned with estimated cash #ows over the whole economic life considered. For plans such as the conventional business budgets, on the other hand, which relate to only a section of the economic-life span of some of the resources concerned, we are compelled to measure pro"t #ows per period. In partial productivity maintenance comparisons between machines, "rms or between divisions of a "rm, just as comparing two maintenance policies o!ering the same productivity index, reference to index alone, may be misleading. In our partial productivity model, the output prices of the produced products and input prices (costs) will change over time. Expected changes in the prices of outputs and of current inputs would be built into the planning model. The ordinary general partial productivity index is the total output divided by one kind of input. Then the amount produced per machinehour is one example of partial productivity, unitbased ratios. In our maintenance case, the maintenance activities (preventive and corrective maintenance) are very heterogenous and a unitbased productivity index is not appropriate, but could for example be the amount produced per maintenance hour.

References
[1] S.T. Cordero, Maintenance Management, Fairmont, Englewood Cli!s, NJ, 1987. [2] J.P. Gilbert, B.F. Finch, Developing maintenance craft labor e$ciency through an integrated planning and control system: A prescriptive model, Journal of Operations Management 6 (1986) 449}459. [3] L.F. Hsu, Optimal preventive maintenance policies in a serial production system, International Journal of Production Research 29 (12) (1991) 2543}2555. [4] C. Sheu, L.J. Krajewski, A decision model for corrective maintenence management, International Journal of Production Research 32 (6) (1994) 1365}1382. [5] E.N. White, Maintenance Planning } Control and Documentation, Gower Press, London, 1973. [6] L.P. Ettkin, D.G. Jahnig, Adapting MRP II for maintenance resource management can provide a strategic advance, Industrial Engineering 18 (1986) 50}59. [7] R.L. Dunn, Maintenance update 88, Plant Engineering 42 (1988) 60}62. [8] T.D. Rishel, D.P. Christy, Incorporating maintenance activities into production planning; integration at the master schedule versus material requirements level, International Journal of Production Research 34 (2) (1996) 421}446. [9] R. Dunn, Advanced maintenance technologies, Plant Engineering 40 (12) (1987) 80}82. [10] S. Hardy, L. Krajewski, A simulation of interactive maintenance decisions, Decision Sciences 6 (1975) 92}105. [11] A. Banerjee, B.B. Flynn, A simulation study of some maintenance policies in a group technology shop, International Journal of Production Research 25 (11) (1987) 1595}1609. [12] M. Boiteux, Re#exions sur la concurrence du rail et de la H route, le declassement des lignes non rentables le de"cit du H chemin de fer, L'Economie electrique, (2), 1955. [13] I.B. Gertsbakh, Models of Preventive Maintenance, North-Holland, Amsterdam, 1977.

58

H. Lofsten/Int. J. Production Economics 63 (2000) 47}58 K [27] H.M. Armitage, A.A. Atkinson, The choice of productivity measures in organizations, in: Kaplan (Ed.), Measures for Manufacturing Excellence, Harvard Business School Press, Boston, MA, 1990. [28] J.L. Mammone, Productivity measurement: A conceptual overview, Management Accounting 2 (1980) 36}42. [29] J. Teague, S. Eilon, Productivity measurement: A brief survey, Applied Economics 5 (1973) 133}145. [30] C. Aggarwal, A study of productivity measures for improving bene"t cost ratios of operating organizations, International Journal of Productivity Research 18 (1) (1980) 83}103. [31] E. Meadows, How three companies increased their productivity, Fortune 101 (1980) 92}101. [32] R.G. Schroeder, J.C. Anderson, G.D. Scudder, White collar productivity measurement, Management Decision 24 (5) (1986) 3}7. [33] M. Babson Jr., Pro"ling your productivity, Management Accounting 4 (1981) 13}17. [34] L.R. Amey, The E$ciency of Business Enterprises, Allen and Unwin, London, 1969. [35] R.L. Acko!, Scienti"c Method, Wiley, New York, 1962. [36] G.J. Staubus, in: D. Richard (Ed.), Activity Costing and Input}Output Accounting, Irwin, Inc, Homewood, IL, 1971. [37] M.J. Farrell, The measurement of productive e$ciency, The Journal of Royal Statistical Society Series A 120 (Part III) (1957) 253}281. [38] M. Brandolese, M. Franci, A. Pozzetti, Production and maintenance integrated planning, International Journal of Production Research 34 (7) (1996) 2059}2075. [39] H. Markowitz, The utility of wealth, Journal of Political Economy 60 (1952) 151}158.

[14] A. Kelly, M.J. Harris, Management of Industrial Maintenance, Newnes Butterworth, London, 1978. [15] C.M. Kelly, C.T. Mosier, F. Mahmood, Impact of maintenance policies on the performance of manufacturing cells, International Journal of Production Research 35 (3) (1997) 767}787. [16] P. Masse, Optimal Investment Decisions, Prentice-Hall, H Inc, Englewood Cli!s, NJ, 1962. [17] B. Naslund, Simultaneous determination of optimal repair K policy and service life, The Swedish Journal of Economics 68 (2) (1966) 63}73. [18] B. Rapp, Models for Optimal Investment and Maintenance Decisions, Almqvist and Wiksell, Stockholm, 1974. [19] H. Hotelling, A general mathematical theory of depreciation, Journal of the American Statistical Association (1925), 340}353. [20] F. Lutz, V. Lutz, The Theory of Investment of the Firm, Princeton University Press, Princeton, NJ, 1951. [21] G. Preinreich, The economic life of industrial equipment, Econometrica 8 (1940) 12}44. [22] H. Smith, Economic equipment policies: An evaluation, Management Science 4 (1957) 20}37. [23] V.L. Smith, Investment and Production: A Study in the Theory of the Capital Using Enterprise, Harvard University Press, Cambridge, MA, 1961. [24] J.S. Taylor, A statistical theory of depreciation, Journal of American Statistical Association 1923, 1010}1023. [25] G. Terborgh, Dynamic Equipment Policy, McGraw Hill, New York, 1949. [26] G. Terborgh, Business Investment Policy, Machine and Allied Products Institute, Washington, DC, 1958.