Handling of production disturbances in the manufacturing industry

2.1 PD in terms of OEE

OEE has been proposed as one way of measuring the losses from PD, and it has been used widely in industry (Kumar et al., 2013). According to the original definition by Nakajima (1988), OEE is a bottom-up approach where an integrated workforce strives to achieve OEE by eliminating the six big losses: equipment failures/breakdowns, set-up/adjustments, idling and minor stoppages, reduced speed, reduced yield, and quality defects. These losses are measured in terms of OEE, which is a function of the three components availability, performance, and quality.

Following this, numerous alternative definitions and additional losses have been proposed. In fact, there has been a gradual expansion of the number of losses to be included in the OEE figure. The original six big losses have been expanded to eight (Ljungberg, 1998) and 11 major factors (Smith and Hawkins, 2004). In addition to the losses captured by OEE, numerous other losses related to labour effectiveness, resource consumption, and safety have been proposed as total preventive maintenance losses (Averill, 2011). Interestingly, a similar development can also be found within Lean, where up to 12 types of wastes have been proposed where in contrast to the original seven (Kyrö et al., 2011). This expansion holds an emphasis on planned activities such as preventive maintenance, shortage of staff from breaks, meetings and education, etc. (Ljungberg, 1998; Jonsson and Lesshammar, 1999; Smith and Hawkins, 2004). In particular, Jonsson and Lesshammar (1999) argue that including planned activities in the production time creates a motive for improvement, e.g., by more efficient set-ups.

Unfortunately, the wide array of OEE definitions has caused misunderstanding and misuse (Williamson, 2006). In an attempt to solve this, the OEE Foundation (2014) has suggested an “OEE Industry Standard” with guidelines on how to define OEE in order to find all potential losses. They state that all losses need to be defined and visualized, and stress that set-ups, breaks, and maintenance are all potential losses to be reduced. In particular, reducing set-up losses through continuous improvements is the core of concepts like Single Minute Exchange of Die (SMED) (Shingo, 1985). Such improvements hold great potential for increasing productivity, made evident by, e.g., Ericsson’s (1997) case studies where 24 per cent of the non-operative machine time consisted of set-ups. Nevertheless, the consistent reports of low-OEE figures illustrate that handling of PDs remains a crucial task at hand, and new OEE definitions and additional losses have not solved the problem. Moreover, despite the extensive use of OEE as a measurement for PDs, there remains a dissent in regard to the understanding of what actually constitutes a PD; in industry and in academia. Overall, this raises the need for a more holistic approach to defining, measuring, and handling of PDs.

2.2 Preventive maintenance as a PD

In terms of OEE, preventive maintenance has been a particularly popular subject of discussion. In Ericsson’s (1997) studies, 39 per cent of non-operative time consisted of maintenance-related activities. Preventive maintenance is typically excluded from OEE as it is assumed you have to do it, you cannot reduce it, and you cannot eliminate it (Smith and Hawkins, 2004). It is also a sort of a paradox: it can cause disruption if carried out during production, even though it is aimed at avoiding failures. In some occasions, preventive maintenance is also a cause of failure due to the probability of introducing failures during the maintenance activity (Ylipää, 2000). In fact, poor scheduling of preventive maintenance also increases the risk of dramatically disturbing production (Wong et al., 2013). Also, one-third of all maintenance costs have been found to be wasted due to unnecessary or improperly carried out maintenance (Mobley, 2002).

According to Katila (2000), the most effective way to reduce technical failures is to increase the amount of preventive maintenance. Therefore, many industrial companies are targeting an 80/20 relationship between preventive and corrective maintenance (Adolfsson and Dahlström, 2011; Wängberg and Larsson, 2013). Sandberg et al. (2014) specifically highlight a case where such a target was set within an organization. However, it was found that the corrective vs preventive ratio was difficult to affect due to several hindering factors, including, e.g., a perception amongst many company functions and employees that preventive maintenance is of no financial value. This issue is, e.g., tackled in reliability-centered maintenance, where a deeper analysis of specific failure modes dictates the chosen maintenance action (i.e. reactive, preventive, or predictive). It is stressed that a preventive maintenance activity should only be chosen if it is effective and applicable, i.e. being both cost-effective and able to detect, prevent, mitigate a failure or discover a hidden failure (Hinchcliffe and Smith, 2003).

Due to the both the high cost of maintenance and its impact on production efficiency, not only a shift towards increased predictive maintenance is emphasized (Hinchcliffe and Smith, 2003). The goal could also be to minimize the maintenance requirement throughout its life cycle by addressing maintainability and reliability at the earliest stage possible, preferably during the design phase of new equipment or machinery (Almgren, 1999; Jonsson, 1999; Ylipää, 2000; Wireman, 2000).

2.3 The use of OEE

Not only has the losses in OEE been widely discussed in literature, but the purpose of OEE has also been debated. It is, for example, argued inappropriate to use this as a plant-level measure or for benchmarking different assets or processes (Williamson, 2006), and more system-oriented versions have therefore been proposed, such as overall line effectiveness (Braglia et al., 2008), and overall throughput effectiveness (Muthiah and Huang, 2007). Another flaw in the original OEE definition is that each of the three components is assumed to be equally important, resulting in a statistically invalid OEE percentage rating. To resolve this, a weighted OEE figure has been proposed (Raouf, 1994; Wudhikarn, 2010; Yuniawan et al., 2013).

However, regardless of the OEE definition or the included losses, striving for maximizing the OEE percentage rating and manipulating calculations to obtain a favourable figure can cause sub-optimization and stimulate overproduction (Van Goubergen, 2010). Therefore, a general consensus exists that the fundamental purpose of OEE is to direct improvements by identifying and reducing specific losses (Nakajima, 1988; Ljungberg, 1998; Jonsson and Lesshammar, 1999; Bamber et al., 2003; Williamson, 2006). Instead of purely focussing on increasing the OEE percentage rating without considering its impact on overall system performance, focus should be on the individual factors where root cause analysis is performed, improvements are identified, and specific losses are eliminated (Smith and Hawkins, 2004; Van Goubergen, 2010).

2.4 Mapping of PD factors

To support the handling of PDs, Ljungberg (1998) argues that it would be beneficial to shift from a narrow focus on down time losses and instead use a more comprehensive model of losses. Moreover, Smet et al. (1997) stress that a necessary condition for successful automation of registering of PDs is an appropriate model of PDs. However, due to the ambiguousness of the definition of a PD, Ingemansson (2004) highlights that the categories of PDs should be clearly distinguished from each other in order to reduce the risk of misconception. Moreover, Ingemansson (2004) believes there can be a debate on what should be regarded as a PD, both in the academic and industrial world. In real life however, it is more important to find a workable definition. Nonetheless, such a definition needs to be able to capture all losses, and PDs therefore need specific classifications for different fields (Smet et al., 1997; Bellgran and Aresu, 2003). Bamber et al. (2003) argue that organizations should develop their own classification frameworks. Depending on the specific field of application, it is not necessarily important that such frameworks explicitly or definitively include the original six big losses in OEE.

Studies highlighting different PD factors have been presented, e.g. in SMEs in New Zealand, where a mapping of 14 operational disturbances that degrade business performance and environment was presented. The authors found that disturbance handling systems are weak and informal (Islam and Tedford, 2012). In another study, in Bangladesh, 11 proposed disturbance factors were mapped, and 27 root causes behind the disturbances were identified. The authors conclude that frequent changeovers in production is one of the most detrimental, and if an organization can systematically identify disturbances and their root causes, overall productivity can easily be improved (Islam et al., 2012).

Another dimension of the handling of PDs is how the individuals who play a central role in achieving high reliability at their company view different PD factors, and how this relates to whether they are being measured or not. Some authors have touched upon this subject, e.g., Ljungberg (1998) who mentions that major time losses such as set-up and adjustments are not regarded as losses but as productive time. However, individual view on various PD factors has not been specifically addressed in literature. A prerequisite for measuring an event as a disturbance is naturally that is must be interpreted as one. Therefore, the fundamental basis for working with individual PD factors in improvement work is to first agree on what factors that should be regarded as PDs and treated as such.

With the assumption that the reader is reasonably well versed in the concept of PDs, the basis for discussion regarding PDs is the definition proposed by Ylipää (2000); “Production disturbances are discrete or decreasing, planned or unplanned, disruptions or change during planned production time, which might affect availability, operational performance, product quality, security, work conditions, environment, etc.” This definition thus includes planned stoppages, as well as any measurable stop, regardless of time limit (in contrast to e.g. Golinska et al., 2011 who define PDs as solely unexpected and unplanned events). The essence is that PDs should be measured as deviations from the normal, desirable conditions. However, this general definition do not explain in detail what constitutes a PD, and the aim of this study is therefore to provide a more detailed view of what factors that can be regarded as PDs within the manufacturing industry.

3.1 Distribution of the questionnaire

Survey A was carried out as part of a national multi-disciplinary co-operation research project called “TIME” during 2001-2004, with researchers from three universities and one industrial research institute (Gullander et al., 2003). The data were collected during May 2001 to August 2001. Parts of the original questionnaire and an extended methodology have been reported in Ingemansson et al. (2002) and Ylipää et al. (2007). This chapter therefore focusses primarily on describing the methodology of survey B.

For survey B, the paper-based questionnaire was reconstructed to a web-based questionnaire, and held open for submission between March and April 2014. Invitation to the questionnaire was sent by e-mail to selected respondents, while an open invitation was listed publicly on the website of Sustainability Circle (SC), and included in an SC e-mail newsletter sent to their subscribers. SC is a non-governmental, maintenance-focussed organization with more than 50 member companies. The companies that participated in survey A received a prior notification call, and an invitation was sent if personal contact with an appropriate representative at the company could be established.

In order to facilitate a survey including different production contexts, a pre-study including five context descriptions of different production systems based on data from 24 respondents from 15 different companies was used to construct the questionnaire in survey A (Ylipää et al., 2007). Due to the transformation from a paper to a web-based questionnaire, pre-testing was performed again. A pilot questionnaire was tested and evaluated by external researchers and industrial experts (Forza, 2002). In order to improve the questionnaire, unstructured telephone interviews followed after completion of the pilot version, and minor revisions were made before airing the final survey.

3.2 Structure of the questionnaire

The questionnaire included questions related to 21 main PD factors. These were originally derived from literature review and case studies performed in a longitudinal research project called CONSENSUS, where critical events and PDs were studied during the running-in of a new product (Harlin, 2000).

Data regarding the 21 PD factors were collected using “yes/no/do not know” alternatives. However, as seen in Tables AI and AII in Appendix, the classification of PDs was asked using a “yes/no” alternative in survey A, and a “yes/no/do not know” alternative in survey B. The registration of PDs was collected using “yes/no/do not know” in both surveys. This inconsistency was due to limitations of the web-based survey tool. As seen in the schematic example in Table I, the respondent first classified the 21 factors as PD or not PD based on personal opinion. Thereafter, they answered whether the factor was measured and registered as PD in the monitoring system at their company. These two questions were asked to gain knowledge about the practical perspective of the handling of PDs, as well as enabling analysis of similarities or differences between the individual views and the registration of PDs. Moreover, the possibility for the respondents to add other PD factors (beyond the 21 proposed factors) was suggested after the completion of survey A. This option was therefore included in survey B by means of an additional open-ended answer alternative.

The remaining parts of the questionnaire were used to explore further aspects concerning the handling of PDs, e.g., the use of engineering tools and methods. The questionnaire also included a final part regarding demographic information about the respondent and the participating company (Table II).

3.3 Selection of companies and respondents

A non-probabilistic judgement sample was chosen in both surveys (Forza, 2002). The intention was to indicate an expert view from a high-strategic level within maintenance or production management at the different companies. This selection was made as these respondents were considered to possess specific knowledge about the production area, and play a central role in achieving high reliability of the production system with respect to PDs.

The sample size of survey A was 80 responses, and the same number was therefore set as a goal for survey B. For survey B, 82 selected respondents received an invitation to answer the questionnaire, out of which 62 answered, resulting in a response rate of 75 per cent. In order in increase the response rate, non-respondents received monitoring phone calls and up to three e-mail reminders. The most common argument for non-response was lack of time. The open invitation resulted in 22 additional responses. Out of the total 84 submissions, the respondents who did not fulfil the criteria of having an expert view from high-strategic level were excluded, and the respondents with the highest management level were chosen at plant-level for each company.

The final selection consisted of 76 responses from 71 companies. The five duplicates represent individual respondents from different plants within the same company or corporate group, but separated geographically and operating with different management. In total, 62 per cent of the respondents can be regarded as maintenance department, 25 per cent production department, and 13 per cent as equally maintenance and production department.

3.4 Data about the respondents and participating companies

A description of the respondents and participating companies in survey A has been reported in Ingemansson et al. (2002). In Table II, the presented categories of company sizes, in terms of number of employees, follow EU classification for enterprise size (Pagell and Halperin, 2000). Turnover is based on 9.52 SEK per 1€. The data illustrate the number of employees, capital turnover per year, and the respondents’ years of experience within production or maintenance. Moreover, regarding industry type, the majority of the respondents, 52 in total, represented discrete manufacturing companies. Thereafter, 11 respondents represented process industry companies, and 13 represented other types of companies, e.g., food, energy, pulp, etc. Several of the participating companies are large multi-national corporations acting on a global market, who may have an internally consistent approach to operations. This implies a possibility that the view of PDs amongst these respondents is also prevalent outside Swedish industry.

3.5 Analysis and presentation of data

In line with the descriptive research approach, the results are presented using descriptive statistics. This is a way of reducing the data to a manageable form (Faber, 2012) and allows for summarizing the results into numbers and graphs that are easy to interpret (Fisher and Marshall, 2009). Discussion of individual PD factors and highlighting of differences are therefore presented in terms of frequency of responses. In order to present the compiled data of the 21 PD factors in a comprehensible way, polar diagrams (also called radar charts, spider diagrams, etc.) are used. This is a graphical tool that allows for a better and more graspable display of multivariate data in comparison to bar charts or text (Saary, 2008).

4.1 Classification of PDs

The polar diagram in Figure 1 (full data in Tables AI and AII in Appendix) displays how individuals in both surveys classify the 21 different PD factors. It shows that 90 per cent or more of the respondents classify the three factors “equipment failure”, “failure in software in production equipment”, and “media error” as PD, i.e. factors that result in clearly visual effects. Moreover, six other factors are classified as PD by 75 per cent or more in both surveys: “human error, “failure of peripheral”, “subsequent stop in output flow”, “speed loss”, “waiting time for incoming material/product”, and “scrap or quality problems”. In contrast, it is observed that planned events such as “tool change” and “set-up/resetting”, “preventive maintenance”, “pauses or breaks”, and “cleaning”, are classified as PDs to a lower extent: by 50 per cent or less in both surveys. This indicates that unplanned events are classified as PDs at a higher degree than planned events.

Moreover, the possibility in survey B to add further PD factors resulted in that seven individuals reported that the following factors are classified and registered as PDs in their company: “errors in drawings”, “OEE”, “process waste”, “delivery service”, “minor stoppages”, and “consumables”. One respondent also classified the factor “negative employees” as a PD, but this was not registered as a PD. Also, one respondent commented that adjustments, set-ups and reprogramming are planned activities and therefore not regarded as PDs.

4.2 Comparing classification of PDs in surveys A and B

The similarity of the two data sets in Figure 1 indicates a general consistency in the classification of PD factors. Further, Table III shows the eight factors with the largest difference between the two surveys. It indicates that “time for refilling/changing material” and “incidents/near misses” are to a greater extent classified as a PD by in survey B, indicated by a difference of 14 absolute per cent. Thereafter, six other factors display a difference of 7-8 absolute per cent. Moreover, by taking the mean value of all 21 factors (i.e. the mean of combining 97 per cent classification of equipment failure as PD, 85 per cent human error, 23 per cent pauses or breaks, etc.), a general mean value of 63 per cent and 65 per cent for the two surveys is found (Tables AI and AII in Appendix). This lends further support to a general consistency of the classification of PDs between the two surveys.

4.3 Registration of PDs

Figure 2 (full data in Tables AI and AII in Appendix) shows to what extent the proposed factors are registered as PDs in the company’s monitoring system. In both surveys, the three factors “equipment failure”, “failure in software in production equipment”, and “media error”, were registered as PDs in approximately 75 per cent or more of the companies. Moreover, four additional factors are registered by 50 per cent or more of the companies in both surveys: “human error”, “failure or peripheral”, “subsequent stop in output flow”, and “scrap or quality problems” as PDs. Planned events are registered at a lower degree: “tool change” and “set-up/resetting” were registered as PDs according to less than 50 per cent of the respondents, and “pauses or breaks” according to less than 25 per cent. Overall, these results are consistent with the classification of PDs, i.e. that unplanned events are emphasized to a larger extent than planned events.

4.4 Comparing registration of PDs in surveys A and B

In contrast to the classification of PD factors, larger differences between the two surveys can be found for the registration of PDs. A general trend that companies in survey B register more of factors as PDs is observed, where in total 14 out of the 21 factors display a difference of 8 to 31 absolute per cent (Table IV). The largest differences are found for “preventive maintenance” and “incidents/near misses”. The companies in survey B clearly register these two factors as PDs to a larger extent, indicated by a difference of 25 and 31 absolute per cent, respectively. Thereafter, “cleaning” and “scrap or quality problems” are also registered to a substantially higher degree in the survey B. The general mean values of which factors are registered as PDs are 45 and 56 per cent for the two surveys, respectively.

4.5 Comparing classification and registration of PDs

A general tendency can be found regarding the difference between classifying and registering the factors as PDs. In both surveys, the individual experts are prone to classify more factors as PDs compared to what is being registered at their company.

In survey A, all 21 factors are at a higher degree classified as PDs, and 13 factors display a difference of 18 absolute per cent or more (Table AI in Appendix). In survey B, ten factors have a difference of 13 absolute per cent or more (Table AII in Appendix). Moreover, when comparing the general mean value of classification and registration, there is a difference of 18 absolute per cent in survey A and 9 absolute per cent in survey B. The largest differences are illustrated in Tables V and VI. In survey A, “planning error” stands out with a difference of 30 absolute per cent. In survey B, the two largest differences are observed for “planning error” (34 absolute per cent) and “shortage of staff” (28 absolute per cent). Interestingly, the factor “preventive maintenance” goes against the trend and is in survey B instead registered as a PD at a higher degree (with a difference of 13 absolute per cent). Overall, these results suggest that there is a gap between the classification and registration of PDs, where the respondents are more prone to classify the proposed factors as PDs compared to what is being registered.

5.1 Interpretation of PD factors

OEE was originally developed to reveal the losses from PD (Nakajima, 1988), but the wide array of definitions have caused misunderstanding and misuse (Williamson, 2006). By measuring individual PDs and the different components of the OEE figure, the OEE can be increased (Ingemansson, 2004). However, a prerequisite for an event to be registered and measured as a PD is naturally that it is interpreted as one. The results from this study show that almost all respondents regard the factors “equipment failure”, “failure of software”, and “media error” as PDs (Figure 1). This goes in line with Ljungberg’s (1998) conclusion that companies focus on down time losses and especially breakdowns. These three factors are also easy to detect due to the visual deviation from the normal state. Furthermore, over 75 per cent of the respondents in both surveys classified eight other factors as PDs. These factors relate to the production flow (“waiting time for incoming product/material”, “subsequent stop in output flow”, and “speed loss”), people (“human error”, “planning error”), and quality (“scrap or quality problems”).

Following Nakajima’s (1988) definition, there has been an increase of the losses proposed for inclusion in OEE. Emphasis has been put on planned activities (Ljungberg, 1998; Jonsson and Lesshammar, 1999). Similarly, Ylipää (2000) includes any planned or unplanned event in his definition of a PD. However, the planned events “tool change” and “set-up/resetting” were regarded as PDs by 50 per cent or less in both surveys (Figure 1). This supports Ljungberg’s (1998) notion that these are often not seen as losses but as productive time, even though frequent changeovers are highly detrimental (Islam et al., 2012) and can make up large parts of a machines non-operative time (Ericsson, 1997). In fact, one respondent even clarified that adjustments, set-ups, and reprogramming are planned activities and therefore not regarded as PDs. Measuring these as losses creates an incentive for improvements, e.g., more efficient set-ups (Jonsson and Lesshammar, 1999). The rather low recognition of these two factors as PDs is noteworthy since they have been included in virtually all OEE definitions since its origin, and have been subjects of continuous improvement in, for example, SMED (Shingo, 1985).

Similarly, other planned events such as “time for changing/refilling of material”, “cleaning”, “work meeting”, and “pauses or breaks” were regarded as PDs by less than 50 per cent of the respondents in both surveys (Figure 1). These events have also been included in OEE definitions (e.g. Ericsson, 1997; Ingemansson, 2004; Smith and Hawkins, 2004). In fact, OEE Foundation (2014) argues that all losses, including planned events, need to be defined and visualized. It shall be stressed that this low recognition of planned events as PDs is based on the view of individual experts with long experience of working at these companies; people within industry who one can expect to possess a high level of competence, knowledge and education within the subject area of PDs.

Due to the ambiguousness of the concept of a PD, Ingemansson (2004) argues that categories of PDs should be clearly distinguished from each other in order to reduce the risk of misconception. This ambiguousness is observed in this study, where, for example, “set-ups” are classified as a PD to a considerably lower extent than “adjustments” in both surveys (Table I). This is rather interesting and surprising since set-ups and adjustments are merged as one of Nakajima’s (1988) original six big losses. An adjustment is however a correction from an unwanted to a satisfactory state, which could explain why this is more often regarded as a PD. In fact, several similar cases of ambiguity are found in this study. For example, whilst 26 per cent or less of the respondents in both surveys classified “pauses or breaks” and “work meetings” as PDs, the rather similar factor “shortage of staff” was classified as a PD by more than 70 per cent in both surveys. Moreover, the possibility of adding PD factors further highlighted how similar events could be classified as different PDs. For example, the additionally mentioned factor “process waste” could potentially be regarded as the proposed factor “scrap or quality problems”. Further, “consumables” (i.e. lack of these) may be interpreted as “waiting time for incoming material”, and “errors in drawings” might be described as a consequence of a “human error”. These examples show how easily different views and multiple definitions and interpretations of PDs can arise.

5.2 Classifying and registering PD factors

From a longitudinal perspective, a general stability can be found regarding how individuals classify the proposed PD factors (Figure 1). The general mean value between the two surveys only differ by 2 absolute per cent, and 19 out of the 21 factors display a mere difference of 8 absolute per cent or less. These results indicate that the view of which factors the individual experts at the companies regard as being PDs have been consistent for over a decade. Interestingly however, the reported OEE figures during the past two decades have been consistently low (Ahlmann, 1993; Ericsson, 1997; Ljungberg, 1998; Ingemansson, 2004; Ylipää et al., forthcoming). This raises the questions of what is actually required in order to achieve effective handling of PDs in industry.

Moreover, a gap between the classification and registration of PDs is found, where individual experts at companies are more prone to classify factors as PDs compared to what is being registered. In survey A, a difference of 18 absolute per cent of the general mean value could be found between classification and registration of PDs, and a difference of 9 absolute per cent was found in survey B. As seen in Tables V and VI, large differences are found for individual factors. These results illustrate how the individual view of managers does not always reflect what is really being measured and registered at the company. Fortunately, this gap seems to be decreasing since this study also observed a trend towards increased registration of PD factors (Figure 2). The general mean value for registration of PD factors is 45 per cent in survey A and 56 per cent in survey B, and individual factors display substantial increase (Table IV). This increase could possibly be an effect of advancements in information technology systems and computerized maintenance management systems. Overall, the results suggest that most companies have started to register various PD factors in their monitoring systems, but have not fully reached a state where all regarded factors are being measured.

5.3 Preventive maintenance as a PD

There has been much debate in literature regarding whether preventive maintenance should be viewed as a PD or not. Typically, it is excluded from OEE calculations as it is assumed you have to do it, you cannot reduce it, and you cannot eliminate it (Smith and Hawkins, 2004). Some authors exclude preventive maintenance from their definitions (Nakajima, 1988; Smet et al., 1997), whilst other emphasize that it should be regarded as a PD and included in OEE (e.g. Ljungberg, 1998; Jonsson and Lesshammar, 1999). However, this study shows that a majority of respondents do not consider the factor “preventive maintenance” to be a PD. Only 32 and 36 per cent of the respondents classify it as a PD in surveys A and B, respectively (Figure 1). Surprisingly however, a positive trend towards increased registration of “preventive maintenance” as a PD is indicated (Figure 2).

In terms of PDs, preventive maintenance is paradoxical in several aspects. First, preventive maintenance is a paradox in itself: even though it is fundamentally aimed at avoiding failures, it can be a disruption if it is carried out during production, and it can also introduce new failures (Ylipää, 2000). Moreover, increasing the amount of preventive maintenance is, according to some authors, the most effective way of preventing technical failures (Katila, 2000). Many industrial companies are also striving to achieve an 80/20 relationship between preventive and corrective maintenance (Adolfsson and Dahlström, 2011; Wängberg and Larsson, 2013). These could be contributing reasons for the increased registration of preventive maintenance as a PD as seen in this study. However, poor scheduling of preventive maintenance can dramatically disturb production (Wong et al., 2013), so the aim of increased preventive maintenance is also a paradox in terms of PDs: reducing potential downtime from failures by increasing preventive maintenance could increase potential downtime from preventive maintenance itself. Also, no studies have shown that 80/20 would be an optimal distribution, and other studies have shown that the corrective vs preventive ratio can be very hard to affect (Sandberg et al., 2014).

Undoubtedly, the complex problems with preventive maintenance are not easy to solve. It is however of great importance, especially since it has been indicated that one-third of total maintenance costs are wasted due to unnecessary or improperly carried out maintenance (Mobley, 2002). This is particularly hard to tackle if preventive maintenance is not even regarded as a PD that could be reduced, an attitude shown to be prevalent according to the data in this study. Clearly, one cannot simply strive to increase preventive maintenance, and in some companies it is not even perceived as financially valuable (Sandberg et al., 2014). It is first when one can collect data about individual PDs and use this as guidance for focussed analysis down to specific failure modes that decisions can be made as to whether preventive maintenance should be increased or not. A systematic framework for classifying, measuring, and registering specific PDs could however help in this work. For example, instead of simply increasing preventive maintenance, there could be a focus on “better” maintenance (e.g. predictive maintenance) or striving to minimize the maintenance requirement by addressing maintainability and reliability during the design phase (Almgren, 1999; Jonsson, 1999; Ylipää, 2000; Wireman, 2000).

5.4 The effect of different views of PDs

The original OEE percentage rating is questioned by, e.g., Liker, Raouf (1994), Wudhikarn (2010), and Yuniawan et al. (2013). However, OEE can be seen as the starting point for identifying PDs and directing further analysis and improvement work, instead of being used as the final measure. In essence, a systematic monitoring of PDs provides a guide for continuous improvements (Smet et al., 1997). However, this study has shown that several of the proposed factors are not regarded as PDs, which arguably omits them from being improved. Within literature, there is support for conflicting views of the concept of PDs between industry, academia, and even within companies and their own organizations (Ljungberg, 1998; Harlin, 2000; Ylipää, 2000). This study has presented further empirical evidence that suggests this scenario to be prevalent, and seems to have been consistent for over a decade. Individuals within industry interpret and classify various PD factors differently, and their views do not always correspond with existing literature. Moreover, the personal view of experienced maintenance and production experts does not always resemble what is being measured and registered.

The combination of consistent reports of low-OEE figures for over two decades, the increasing number of OEE definitions and proposed losses, and the differences in the view of PDs shown in this study, clearly highlight the need for more detailed and structured frameworks for handling of PDs. The consequence of classifying PDs differently is that they will be resolved differently (Ylipää, 2000), and this is the fundamental practical issue to overcome from a managerial perspective. Therefore, organization must establish a common, internal understanding of what events that should be regarded as PDs and treated as such. This understanding should constitute the basis for developing their own classification framework that corresponds to the specific field of business (Bellgran and Aresu, 2003; Bamber et al., 2003). By doing so, an organization can systematically identify PDs and their associated root causes, and overall productivity can thus be improved (Islam et al., 2012).

Finally, the urgency of this matter will increase in line with the development towards digital manufacturing. In the digital factories of the future, companies will be able to collect enormous amounts of data regarding potential PDs, and analyses of such data will be the primary means for identifying, reducing, and eliminating PDs. Since a necessary condition for successful automation of registering of PDs is an appropriate model of PDs (Smet et al., 1997), industrial managers must work towards achieving an internal consensus of what events that should be classified and registered as PDs.

5.5 Methodology discussion

Naturally, the potential PD factors included in this study are not exhaustive. In fact, the additional comments in survey B illustrated that other factors can absolutely be considered as PDs. Nonetheless, the list of 21 factors is seen as being a list of key PDs factors in the manufacturing industry, which should be completed in relation to specific industrial or individual contexts.

This longitudinal study involves a fairly large sample including both SMEs and large multi-national corporations based on a non-random sampling strategy, which limits the overall generalizability of the results. Moreover, the targeted sample for the open invitation could not be quantified, thus prohibiting the possibility of establishing a definitive response rate. These responses could also potentially be influenced by self-selection bias (Forza, 2002). However, the fact that responses from individuals not having an expert view were excluded ensured that the open invitation resulted in relevant data. Further, the responses are based on an expert view on a high-strategic level, a group that is often hard to obtain in questionnaire studies. Therefore, the authors are convinced that this study provides valuable empirical data that contribute to an increased understanding of how various PDs are classified and registered in industry. Moreover, the intention of the study has not been to establish any exact numbers of any PD factors. Instead, it aims at achieving a descriptive picture of how PDs are regarded in Swedish industry.

5.6 Suggestions for further research

One aspect to consider is the potential differences in the view of PDs between organizational levels. This study was based on the views of maintenance or production experts at high level. If there are such widespread views amongst management, there are likely to be even larger differences at operational level. If OEE, or handling of PDs in general, is to be the bottom-up approach that it intends to be, a shared view of what factors should be regarded as PDs needs to be established throughout the organization. Therefore, a suggestion would be to replicate this study with a sample of shop floor workers.

In order to fully operationalize the concept of PDs and a factor-based approach to handling them, further research towards data-driven analysis on an individual PD level is needed. These analyses could be used for immediate maintenance actions as well as input to engineering activities aimed at a reduction of production losses. This also requires further development of monitoring systems that are based on established PD classification frameworks. Furthermore, a fundamental prerequisite for using collection and analysis of production data as a mean for identifying, reducing, and eliminating PDs, is the development of systematic approaches that ensures the quality of production data.