Designing a practical asset intervention policy based on product quality degradation

(1)

Designing a practical asset intervention policy based

on product quality degradation

- Quality means doing it right when no one is looking –

Henry Ford

University of Groningen

Faculty of Economics and Business

MSC. Technology and Operations Management

Author: Mark Terpstra

S1913522

m.terpstra.7@student.rug.nl

(2)

Abstract

Purpose

The maintenance function within organisations is evolving from a necessary evil towards a contributor to profit. The fix-it-when-it-breaks perspective is gradually replaced with a view that the maintenance function should be focused on preventing failures and improving the process. Together with this development the quality requirements of customers are only increasing. Although the maintenance perception is changing and the quality requirements of customers are increasing, the maintenance activities in practice are still predominantly corrective. Apparently it is hard for practice to implement proper preventive maintenance policies. Both for the purpose of extending the literature and for adding to the knowledge of practitioners it would be beneficial to gain more insight into the factors influencing the maintenance decision. Moreover, learning more about the influence of maintenance on product quality helps identifying opportunities for improvement. Therefore the goal of this research is to answer the following question: How can a decision making model be designed that supports the

selection of the right maintenance policy based on product quality degradation?

Methodology

In this research the design science research methodology is used for designing the decision-making model. This method is used as it is driven by field problems and its core product is a generic design. The data is gathered through a literature review and a case study at a metal packaging company. The results from both are then used to create a generic decision-making model.

Findings

(3)

Preface

This thesis is my final step on the educational journey I followed at the University of Groningen. It concludes my master’s program Technology and Operations Management and challenged me in every possible way. The aim of this thesis is to narrow the gap between maintenance theory and practice and to expand the knowledge on maintenance management for both academics and practitioners.

In order to do so it is important that the opportunity is offered to conduct research in practice and therefore I would like to thank the case company for giving me this opportunity. I would like to thank Paul Trip for making this research possible and René Keizer for introducing me within the company and supporting me during this project. I would also like to thank everyone else at the company who cooperated with me during my research.

From the University of Groningen I would like to thank my supervisor dr. ir. Wilfred Alsem for providing me with constructive feedback and helping me keeping the project on track. Also his efforts to help me get started at the beginning of the research are highly appreciated. I would also like to thank my co-supervisor Nick Ziengs. Especially during the proposal phase for providing me with valuable feedback and directions.

Finally I would like to thank my family for supporting me during this though journey, my father for all the welcome coffee breaks, and off course my girlfriend for supporting and motivating me at home. Hoogeveen, April 2017.

(4)

1. Introduction

In current global and competitive markets, equipment needs to be efficient, be as reliable as possible and produce high quality products. Especially this last factor is important as only good products provide revenue for the company. This means that during the lifetime of an asset it needs to be properly taken care of to prevent the assets from failing. Although this seems like an open door, past research has provided us with multiple maintenance strategies and policies and with different methods for measuring the effectiveness of the maintenance function. Selecting the appropriate maintenance strategy is of upmost importance to prevent assets from failing.

Maintenance can be divided into two main categories: corrective and preventive maintenance. Preventive maintenance can be divided into time-based maintenance and condition-based maintenance, in which the latter is a rather new approach to preventive maintenance (Prajapati, Bechtel, & Ganesan, 2012). Selecting the appropriate type of maintenance is a difficult task but some past researches have attempted to distinguish these types by setting certain criteria. Waeyenbergh & Pintelon, (2002), proposed a decision tree that selects the appropriate maintenance policy based on a few simple questions. However when a certain maintenance policy is selected by such a model, the exact content and boundaries of this policy still needs to be determined.

That the definition of maintenance differs is supported by Tsang, (1999), who states that the maintenance function can be seen in a narrow fashion as to fix broken items, but also in a wider context in which it covers every stage in the life cycle of equipment. It is however important to note that as the definition of maintenance becomes wider, it is getting increasingly more difficult to distinguish between maintenance activities and operational activities. One of these wider approaches is total productive maintenance (TPM) which is focussed at maximizing equipment effectiveness. In order to measure the effectiveness of this approach the overall equipment effectiveness (OEE) metric is proposed, which is interpreted as the multiplication of availability, performance and quality (Jeong and Phillips, 2001). This metric is one of the most common used metrics in practice to determine the effectiveness of the maintenance function.

(7)

time-based, initiated by certain thresholds, or based on equipment or product degradation. Which initiation point is taken is strongly related to the goal of the maintenance function. When controlling cost of the maintenance function is most important, planned maintenance may be the most appropriate policy. When the quality of the end product is most important, it may be the most appropriate to monitor product quality degradation and to interfere in the process when the quality of the product runs out of specification. However, the question remains how such a policy should be implemented and whether such a policy can prevent gradual degradation of products.

Although a large amount of researches exist on maintenance strategies, performance indicators, or initiation points, generalizable models or frameworks for integrating or implementing maintenance policies hardly exist. This research attempts to provide such an integrated framework that shows the set of interventions necessary to design a maintenance policy based on product quality degradation. The goal of this research the following question:

How can a decision making model be designed that supports the selection of the right maintenance policy based on product quality degradation?

This research uses the design science methodology to come up with a maintenance decision model that can be used in different situations. The data for this study will be gathered through a literature review and research at a case company: A metal packaging company. This company operates in the very quality sensitive market of infant nutrition. It therefore makes sense that such companies need to interfere with the process as soon as possible when quality issues occur and need to continuously improve the process to prevent such issues in the future. This makes this company very relevant for this research.

(8)

2. Theoretical Background

The aim of this chapter is to provide an extensive overview of existing literature on the research subject. This gives the reader an indication of the current state-of-art in literature and provides the researcher with an opportunity to indicate possible gaps in literature. According to Thomé, Scavarda, & Scavarda, (2016), the first step in conducting a literature review is determining the scope of the research.

2.1 Scope

Identifying the scope of a literature review is important as it outsets the aim and reach of the review, directs the research, and brings structure in reporting of the results (Thomé, Scavarda, & Scavarda, 2016). The scope of the review consists of research focus, goal, perspective, coverage, organisation and audience. Each of these elements will now be discussed.

In this review the focus is on maintenance and more specifically on: the different maintenance strategies and policies, maintenance decision models, factors that influence the maintenance decision, and the relation between maintenance and quality. The goal is to identify central issues in the research field and to identify gaps in literature that this thesis can contribute to. The researcher has a neutral perspective as the intention is not to critique or defend a certain position but to provide a comprehensive overview of existing literature related to maintenance phenomena. This means that the purpose of this review is to gather enough information to describe and understand the relevant factors and/or phenomena and not to provide an overview of all existing literature related to the topics. This review is organized in a conceptual manner to identify the concepts that influence and determine the maintenance function. Relations between concepts are sought after and presented in a conceptual model at the end of this chapter. This thesis attempts to narrow the gap between maintenance theory and practice and is intended for scholars, practitioners and the general public.

(9)

2.2 Review

The review begins with the definition of the maintenance function itself. After that the different types of maintenance are discussed. Then a selection of maintenance decision models is presented. This section is followed by the factors that influence the maintenance decision. After that the relation between maintenance and quality is discussed and the review ends with some insights into maintenance performance measurement.

2.2.1 Maintenance function

As long as equipment is involved in manufacturing, there will be the need for maintenance. This relation exists because all equipment is under the influence of some degree of degradation which causes the equipment to malfunction or to break down. As the customer’s belief in the product can be seriously damaged by such a breakdown, improving maintenance for preventing failures or degradation has priority over any other things in a company (Shin & Jun, 2015). This section starts with defining the concept of maintenance and will further elaborate on the different aspects of maintenance. The traditional role of maintenance was to fix broken items (Tsang, Jardine, & Kolodny, 1999). With time however, maintenance became more important and according to Pintelon, Pinjala, & Vereecke, (2006), it is an essential business support function that should proactively contribute to the competitive advantage of a company. Within this research, maintenance is defined according to the definition used by Alsyouf, 2009; Dekker, 1996, which originates from the British standard glossary of terms used in terotechnology 1993. This standard defines maintenance as:

“The combination of all technical and associated administrative actions intended to retain an item in, or restore it to, a state at which it can perform its required function.”

(10)

Figure 1: Development of maintenance policy (Arunraj & Maiti, 2007).

Underlying the maintenance discipline, and therefore also these philosophies, are the following concepts: Reliability centered maintenance (RCM), total productive maintenance (TPM), business centered maintenance (BCM), and total quality maintenance (TQM) (Maletič, Maletič, Al-Najjar, & Gomišček, 2014). In their research, Waeyenbergh & Pintelon, (2002) describe RCM, BCM, TPM, and Life Cycle Cost-approaches (LCC) as the most important maintenance concepts. In general it can be concluded that RCM and TPM are the most well-known and used maintenance concepts. RCM was originally intended for maintaining aircraft and aims at directing the maintenance efforts to the parts of the system where reliability is critical (Dekker, 1996). RCM is a costly and lengthy process and may be only justifiable for complex and high-risk systems (Tsang, 2002). TPM aims at identifying and eliminating the six losses in order to maximize the overall equipment effectiveness (OEE) (Duffuaa & Raouf, 2015). Although these concepts are fundamentally different, the maintenance policies applied can still either be corrective or preventive. The following sections will discuss both policies in more detail.

2.2.2 Corrective maintenance

(11)

In this research, corrective maintenance is defined as:

“Action undertaken to restore the functional capabilities of failed or malfunctioned equipment or systems.”

The reactive part of this definition is the fact that the maintenance action is triggered by the unscheduled event of an equipment failure. Tsang (1995) also indicates that this kind of maintenance is often related to high maintenance cost because of the high cost of restoring equipment, the secondary damage and safety/health hazards, and the possible penalty associated with lost production. Corrective maintenance can be divided in immediate and deferred maintenance (Stenström et al., 2015). This distinction seems logical as for example resources for immediate repair are not available and repairing actions need to be postponed.

2.2.3 Preventive maintenance

Instead of performing maintenance after a failure or breakdown has occurred, most recent attention goes out to preventing such an event. The reasons for preventive maintenance being the preferred approach are: reducing the frequency of failures through, for example, lubrication and cleaning, mitigate the effect of failures, warning of impending failure, costs of an emergency breakdown are higher than a planned one (Duffuaa & Raouf, 2015). There are different policies and models available for organizing the preventive maintenance function. Before examining these different policies, first preventive maintenance itself is defined. As with corrective maintenance there are several different definitions of preventive maintenance proposed in literature. The common point in these definitions is that maintenance is performed proactive before the state of the equipment becomes unacceptable. Within this research, preventive maintenance is defined according to the definition used by Stenström et al., (2015), which originates from the European committee of standardization EN 13306. This standard defines preventive maintenance as:

“Maintenance carried out at predetermined intervals or according to prescribed criteria and is intended to reduce the probability of failure or the degradation of items.”

(12)

Types of preventive maintenance

There are several different approaches on preventive maintenance and for each of them there is a wide variety of models available in literature. Kim, Ahn, & Yeo, (2016) state that there are two distinct preventive maintenance policies available: Time-based maintenance (TBM) and Condition-based maintenance (CBM).

TBM

Barlow & Hunter, (1960) were among the first to develop policies for TBM. According to the authors there are two different TBM policies. The first policy is to perform preventive maintenance after a fixed amount of hours of continuing operation without failure. The second policy is to perform preventive maintenance after the system has been operating for a total amount of hours without considering the amount of intervening failures. This means that a part is replaced at failure or at a fixed time. This type of maintenance is also referred to as age-dependent preventive maintenance. With this policy, maintenance is only performed at a predetermined age or at failure (Wang, 2002). Next to age-dependent PM there is the periodic preventive maintenance policy. Under this policy preventive maintenance is performed at fixed time intervals independent of the failure history of the unit (Wang, 2002). For example, management could decide to perform preventive maintenance once a year. When such an interval is set a periodic preventive maintenance schedule is used.

Numerous models for time-based maintenance have been developed in recent years. Analytical and mathematical models were developed to determine optimal maintenance intervals with the use of different distributions. Many of these models were aimed at minimizing total cost. For a review of these models the reader is referred to Wang (2002) and Kim, Ahn, & Yeo, (2016).

CBM

Condition-based maintenance has received much attention in recent years and the body of knowledge is consistently growing. It is an approach that combines data-driven reliability models with sensor data and aims at reducing unnecessary maintenance action (Alaswad & Xiang, 2017). In this research, condition-based maintenance is defined according to the definition provided by Shin & Jun, (2015):

(13)

From this definition, the most important aspect of CBM can be derived namely that CBM warns us before failure happens and gives us increased precision in failure prediction. As the goal of CBM is to identify failures before they actually happen, an important principle of CBM is the P-F curve. This curve, presented in figure 2, shows a failure pattern and identifies the point at which the failure starts, the point at which the failure can be detected and the point where the part is failed. The time between detection of failure and the actual failure can be seen as an opportunity window for eliminating the failure or to mitigate its effect (Bousdekis, Magoutas, Apostolou, & Mentzas, 2015).

Tsang (1995) states that there are three types of decisions involved in condition-based maintenance: Selecting the parameters to be monitored, determining the inspection frequency, and establishing the warning limit. Monitoring the parameters can be done in several different ways of which censoring and inspection are the most important. Censoring means continuous monitoring of parameters to provide real time information on the state of the system. However such monitoring systems might be very costly and could lead to unnecessary maintenance actions (Alaswad & Xiang, 2017). The other option is periodic inspections on parameters or equipment. The results of these inspections provide the input necessary for determining the appropriate maintenance activities (Kim, Ahn, & Yeo, 2016) Machado & Haskins (2016) state that the main problem with the models that are currently available concern data collection and analysis. The suggestion is made that in order to design and operate an optimized maintenance program, the following preconditions need to hold: improve the quality usage of data, perform criticality analysis, establish monitoring and feedback routines, improve, if necessary, the related work processes, and install an appropriate analytical capacity.

(14)

Peng, Dong, & Zuo, (2010) state that two important aspects of a CBM program are diagnostics and prognostics. Diagnostics is concerned with fault detection while prognostics are aimed at predicting faults and degradation before they occur. The authors divide the prognostic models into four categories: The physical model, the knowledge-based model, the data-driven model, and the combination model. The physical model uses mathematical modelling, the knowledge-based model uses expert systems or fuzzy logic and the data-driven model is based upon statistical and learning techniques. The combination model combines aspects of the before mentioned models with the aim to build a more comprehensive model that captures a more complete image of the phenomenon.

One last maintenance policy that deserves some attention is the opportunistic maintenance policy. Maintenance actions are performed on a given part when the state of the rest of the system indicates that maintenance is required. If maintenance is performed on a failed part, the system is checked for other parts that have the potential to fail in the near future. These parts are then also repaired or replaced (Ab-Samat, & Kamaruddin, 2014).

2.2.4 Multiple maintenance policies

(15)

After discussing the different maintenance philosophies and policies, figure 3 presents these in a schematic overview.

These policies differ in the way the maintenance function is organized and which parameters are used as the foundation for the maintenance activities. Corrective maintenance is solely focused on restoring equipment after a breakdown has occurred which is a costly strategy and does not contribute to improving either the performance of the production process or the quality of the output. The preventive maintenance policies ask for a more structured development of the maintenance function but are more aimed at improving the performance of the process and the quality of the output. Time-based maintenance requires machine failure data to come up with a solid maintenance schedule and is therefore more aimed at improving the system performance. However the general idea of performing maintenance is to make sure that equipment produces the output as it is intended. Therefore, maintaining equipment to keep it in the required state contributes indirectly to the quality of the end product. When looking at the quality of the product it seems that, after investigating the different policies, condition-based maintenance is the most appropriate policy. On the other side, CBM can also be a very expensive policy which might not be suitable for all systems. Gathering and analyzing the data might be difficult due to large amounts of parameters and inspection or monitoring errors and requires a lot of effort for successful implementation

(16)

2.2.5 Maintenance management models

Several different approaches/models/frameworks for maintenance organization and management are available in current research. This section shows a selection of these models and the importance for this research is highlighted. It is these models that form the foundation for the decision-model proposed in this study.

Duffuaa & Campbell, (2015) state that proper asset management has an effective maintenance operation and control system as its backbone. Figure 4 shows the maintenance control cycle proposed by the authors. The concepts presented in this control cycle require standardized administering procedures, standardized data collection and analysis, and the ability to effectively report information.

This model shows which mechanisms need to be in place in order to establish a solid maintenance program. It can therefore be used as a template to see whether all aspects within a proposed maintenance program are covered. Important for this research is that this maintenance control cycle incorporates the quality level as one of the two objectives of the plant output. Moreover, it uses quality reports as part of the control cycle to determine the need for adapting the maintenance activities.

(17)

Waeyenbergh & Pintelon (2002) were among the first to recognize that a maintenance concept should be customized to be tailored to the needs of the company, but that the underlying structure of such a concept may be comparable. The framework proposed by these authors consists of the following steps: Start-up and identification of objectives and resources, identification of the most important systems, criticality analysis, maintenance policy decision step, optimization of the preventive maintenance policy, and performance measurement and continuous improvement. Figure 5 shows the framework schematically. In addition to this framework the same authors also proposed a maintenance policy decision tree. Based on decision whether a certain policy is technically or economically feasible, the appropriate maintenance policy is suggested. Figure 6 shows the proposed maintenance policy decision tree.

(18)

Another interesting and rather extensive model is the maintenance improvement model based on downtime analysis (Koussaimi, Bouami, & Elfezazi, 2016). This model identifies problem areas based on abnormal downtimes. The FMECA procedure is than applied to these problem areas and all possible failure causes are identified. Root cause analysis is then performed and possible solutions to the problem are identified. Figure 7 shows the proposed approach. The model requires a well-built downtime database as it otherwise becomes impossible to identify problem areas.

(19)

Maintenance method

Before the maintenance activities can be managed properly, it is necessary to determine which parts need a certain maintenance policy and how the data is going to be collected. Several different maintenance methods have been developed in past research. In their literature review, Ding & Kamaruddin, (2014) made a classification of the different maintenance methods based on the information available to a company. Figures 8 & 9 show the information model and the classification model that the authors proposed in their study. Some of these different methods will be shortly discussed in this section.

(20)

Graphical method

Graphical modelling is a rather simple technique that does not require complicated optimization procedures (Ding & Kamaruddin, 2014). The method identifies the most problematic machines by ordering them according to both downtime and frequency of maintenance need. This method requires accurate downtime registration but the advantage is the fast indication of problem areas. After assigning a high, medium, or low value for both criteria, the machines are placed on a decision making grid. This grid shows which maintenance policy is best suited for each machine (Labib, 1998). Figure 10 shows this decision making grid.

Figure 9: Certainty theory continuum (Ding & Kamaruddin, 2014)

Figure 8: Classification of maintenance policy optimization model (Ding & Kamaruddin, 2014)

(21)

FMEA/FMECA

Failure Mode and Effects Analysis is a tool that is predominantly used by safety and reliability engineers to identify those system elements that could lead to failures with undesirable outcomes (Sharma & Sharma, 2010). When this technique is used for criticality analysis it is referred to as failure mode, effects and criticality analysis (FMECA) (Liu, Chen, You, & Li, 2014). By means of identifying possible root causes and failure modes, the estimated relative risk of those systems is determined and actions are undertaken to limit or avoid those risks (Arabian-Hoseynabadi, Oraee, & Tavner, 2010). These failure modes are ways in which an asset can fail and for each failure mode the severity, probability and risk of non-detection are estimated (Braaksma, Klingenberg, & Veldman, 2013). These values can then be combined and multiplied to form the risk priority number (RPN). The elements with the highest RPN are considered most risky and can be addressed first (Arabian-Hoseynabadi, Oraee, & Tavner, 2010). Although FMEA is a very popular tool it is also generally criticized for two reasons. The first is that various values for severity, probability and detectability might produce an identical RPN while the risk implications might be totally different. The second reason is that the method does not incorporate the relative importance of the three factors. It is assumed that all factors are equally important but in practice this is rarely the case (Sharma & Sharma, 2010). Some authors have tried to overcome these shortcomings by applying fuzzy set theory to the FMEA approach (Liu, Chen, You, & Li, 2014)

2.2.6 Factors influencing the maintenance decision

As mentioned in the introduction section of this thesis, it proves to be quite a difficult task to implement and manage a proper maintenance policy. Reason for this is the wide variety of factors that can influence the maintenance decision. This section presents an overview of these factors found in existing research.

Perfect/Imperfect maintenance

(22)

of imperfect maintenance is taking a more dominant role. Imperfect maintenance is defined as an activity that brings the system in a state between as good as new and as bad as old (Khatab, Ait-Kadi, & Rezg, 2013). Several models are used to model imperfect maintenance, but virtual age models are most frequently used. These models determine the wear-out not on the chronological age, but on a virtual age between zero and the elapsed time since the system was new (Nguyen, Dijoux, & Fouladirad, 2017).

Data requirements

Data is one of the most important requirements for developing a maintenance plan, but gathering the data is one of the most difficult jobs (Waeyenbergh & Pintelon, 2002). The same authors make a distinction between first level data and second level data. First level data can be acquired through suppliers but second level data is in de minds of people, on paper or in computer files. Gathering this data takes a considerable amount of time and work and is therefore one of the reasons that most maintenance concepts are too hard to be implemented on company level. The opposite is proposed by Galar, Parida, Stenström, & Berges (2013) who state that nowadays with the use of powerful hardware systems and software, it has become relatively easy and cheap to acquire data. This makes that data overload is an issue nowadays.

Inspection/sampling schedule

An inspection schedule is part of any preventive maintenance program. The objective of inspection is to detect warning signals that show that a failure is upcoming. The issue with inspections is that it generally relies on human judgement which brings subjectivity in data collection and interpretation (Duffuaa & Raouf, 2015). As with the maintenance policies, a lot of research has been performed on inspection schedules. Yin, Zhang, Zhu, Deng, & He, (2015) for example propose a schedule that uses a delayed monitoring policy. The authors state that because of the low chance of equipment failure or quality shifts at an early age of the production process, the inspection frequency needs to be reconsidered. .

Human interference

(23)

human errors, and that new maintenance strategies should take these errors into account. Alaswad & Xiang, (2017) state that human errors are also a risk in inspection policies.

2.2.7 The link between maintenance and quality

In their book, Duffuaa & Raouf, (2015) state that “Organizations should strive to tie their maintenance activities to the quality of their products”. This statement implies that not all maintenance activities necessarily are aimed at improving product quality. This in contrast to what was mentioned earlier that maintenance improves the reliability and availability of the system ensuring the delivery of high quality products (Ding & Kamaruddin, 2014). Apparently there are different opinions on this matter and therefore the relation between maintenance and quality is further discussed in this section.

(24)

degradation process of products within batches. They say that each batch produced is off less quality compared to the previous batch. The authors propose two different strategies. The first in which the non-conforming items are sold at a discount price and the second in which the products are al reworked in order to obtain the highest selling price. Two mathematical models are designed in order to select the appropriate batch size with the objective to maximize profit. Kurniati, Yeh, & Lin, (2015) link maintenance and quality by means of quality inspection. The authors propose a framework in which the quality inspection takes a leading role within the proces. The inspection activity becomes the trigger for the demand of maintenance. This means that instead of analyzing historical data of equipment to base the maintenance decisions upon, the data from quality inspection is used for this decision proces. Yeung, Cassady, & Schneider, (2007) created an algorithm that finds the near-optimal combination of PM and Statistical Process Control (SPC) under an economic objective. The results of their studies showed that managers should focus on improving the mean time to failure and on improving the detectability of possible shifts in the process. Another interesting find was that the impact of producing products of inferior quality on the overall production cost is very limited. Therefore it is not neccesary for management to devote extensive efforts on this area. Bouslah, Gharbi, & Pellerin, (2016) propose a model that jointly designs and optimizes the production, quality control and maintenance policies with the aim of minimizing total cost while meeting the average outgoing quality limit constraint.

2.2.8 Performance measurement

(25)

2.3 Conclusion

This chapter has investigated the different maintenance policies, management models and methods, and the link between maintenance and the quality of the end product. Each maintenance policy or strategy has its unique characteristics and implementation issues. It was noticed that although a certain maintenance policy can be used as a general view on how the maintenance function should be organised, the strict use of the maintenance policy is only appropriate in specific situations or with specific systems. In the light of quality this is an important aspect as one could argue that not all parts of the system are equally relevant for the quality of the output. So in situations where the quality of the output is most important, it makes sense to combine different maintenance policies and to apply the different policies to specific parts of the system.

(26)

3. Research design

3.1 Problem identification 3.1.1 Company profile

The case company is part of a large international company with 110 glass and metal manufacturing facilities in 22 countries. The company manufactures metal packaging solutions for the nutrition industry. This is one of the most demanding markets for packaging manufacturers as it requires very high hygiene and quality standards. Their core business is the manufacturing and/or printing of packaging solutions. The plant consists of the following three departments: Printshop, Ends, and Cans. The printshop applies color images and clear coat on sheet metal plates, the ends department manufactures different types of lids, and the cans department manufactures cans from the sheet metal provided by the printshop and assembles the lids. This research is carried out at the Ends department which produces lids in different diameters and with different opening solutions. In order to keep the research manageable and to be able to incorporate as much detail as possible, the research is focused on a single production line. More information on this production line is provided in section 3.1.3 and in appendix 2.

3.1.2 Company problem

(27)

equipment. Although these activities are planned, the schedule is not always leading. In case of high demand or rush orders it happens that maintenance is postponed. During this research it occurred that due to production problems, inventory was dramatically reduced and the decision was made to postpone the preventive maintenance in order to prevent any more loss of capacity.

Recent client complaints regarding the quality of the product gave the company some serious issues and renewed attention is directed towards the maintenance function. The management has indicated that it wants to move to a more preventive strategy in which the quality of the end product is leading and should be the overall focus. Therefore the company came with the following question:

How can we move to a more preventive maintenance policy in which the quality of the final product is leading?

3.1.3 Unit of analysis

The unit of analysis in this study is a single production line consisting of several different machines. The machines are coupled in series with only one decoupling point which is located after the press. The entire line can be divided into two tracks A & B and each track has its dedicated machines. The line produces 99mm diameter ends. Figure 12 shows the production process schematically and in the appendix, a more detailed layout of the production line is presented.

3.1.4 Problem Relevance

The wide variety of available maintenance models, strategies, and policies shows the difficulty of implementing an appropriate maintenance function. This research tries to narrow this gap by finding out which factors influence the maintenance function and by proposing a model that companies can use as a guideline for implementing an appropriate maintenance policy. In order to do so the company question is first rephrased into a more general research question:

(28)

In order to provide a solid answer to this main research question, the following questions need to be answered by this research:

- What maintenance policies exist and what are the boundary conditions for these policies? - What maintenance-decision models are available and how can these be applied in practice? - What factors can influence the maintenance decision?

- How should the maintenance function deal with these implementation factors?

The first and second question will be answered by the literature review performed in the previous chapter. The third and fourth question will be answered by both the literature review and the research performed at the case company. The results section is structured according to the questions and the answers are presented there.

3.2 Methodology

In operations management research there exists a gap between research and practice and there is the urge to make operations management research more relevant (Tang, 2016). To do so, Tang also states that because operations management is an applied research discipline, practical relevance should be required in this kind of research. Hill, Nicholson, & Westbrook, (1999) identify the need for new research methods because operations managers have to deal with unstructured problems that must be managed and cannot be modelled. Therefore the traditional techniques like simulation and mathematical modelling are unlikely to be suitable in such situations.

(29)

The design science methodology consists of the following six steps: problem identification and motivation, definition of objectives, design and development, demonstration, evaluation, and communication (Peffers, Tuunanen, Rothenberger, & Chatterjee, 2007). The following chapters are structured according to these steps. The product of design science research is a viable artefact that can have the form of a construct, a model, a method or an instantiation (Hevner, March, Park, & Ram, 2004). The product of this research is a generic model that shows the opportunities and the limitations of certain maintenance policies, and by means of a flowchart the steps to an appropriate maintenance policy are presented.

3.2.1 Data collection

Data collection within case research can be done in multiple ways. According to Voss, Tsikriktsis, & Frohlich, (2002) the prime source of data is structured interviews which could be supported by unstructured interviews or interactions. Observations, informal conversations, attendance at meetings, or review of archival sources are other examples of data sources. McCutcheon & Meredith, (1993) divide collected data into two types, primary data and secondary data. How both types of data will be acquired in this research is shown next.

Primary data

Primary data comes through direct observations of the researcher or by interviewing people that are involved ((McCutcheon & Meredith, 1993). In this research data is primarily gathered trough semi-structured interviews. Reason for the semi-semi-structured approach is that the researcher is interested in all factors that can influence the maintenance decision. As the researcher cannot know all these factors beforehand, the purpose of this interview structure is to get a comprehensive conversation that allows subjects to be brought in by the interviewee. In addition to interviews data was gathered through observations, informal conversations, and by attending morning meetings.

Secondary data

(30)

As became clear from the literature review, there is a wide variety of aspects that need to be taken into consideration when designing or implementing a maintenance policy. As this research is aimed at providing or designing a framework, the data collection is focussed at understanding the underlying mechanisms and issues. The research is not aimed at providing a detailed maintenance schedule that can be implemented at the specific company. This means that the most important data for this research are the interviews held with all the employees that are involved in the maintenance function. The data derived from presentations or documents can be valuable to understand whether the maintenance is performed according to predetermined procedures or agreements.

3.2 Conceptual Model

Figure 13 shows the conceptual model of this research. The purpose of this model is to clearly show the research boundaries and to provide a clear picture to the reader of how this research is organized. The theoretical framework is constructed by a comprehensive review of the literature focussed on the different, and most recent, maintenance management models and maintenance policies. The theoretical framework concludes with an exhaustive overview of all factors that can possibly influence the maintenance decision. It needs to be noted that these factors need to apply to either the decision-making process or the machine condition. Spare-parts inventory, maintenance resource planning and issues like such are outside the scope of this research. Together with the data acquired from the case company, a maintenance decision-making model is proposed based on the design science methodology.

(31)

4. Results

After the problem identification, the definition of objectives and design and development stages begin (Peffers, Tuunanen, Rothenberger, & Chatterjee, 2007). The information required for these steps in the design science methodology is obtained through the research performed at the case company. This chapter presents the results of this research.

4.1 Definition of objectives

The objectives for a solution should result rationally from the problem specification (Peffers, Tuunanen, Rothenberger, & Chatterjee, 2007). The problem presented by the plant manager and the department manager is that the production process is not reliable enough and that there are too many corrective maintenance activities. As a consequence there are a lot of quality issues and the company wants to move to a more preventive maintenance policy. Therefore the objective of this study is to provide a model that supports the maintenance decision making process. In this section all the information that was gathered at the case company is presented. This section will also show whether all employees have the same objective regarding the maintenance function or if this differs between employees.

The first sources of data were interviews with several employees that are involved in the maintenance function. As this research aims at identifying factors that influence both the maintenance decision as the machine condition, people from different ranks and positions were interviewed. Six interviews were conducted with the department manager, team leaders, process engineer, technical assistant and process technician. These different positions will be shortly discussed.

The department manager is responsible for the overall performance of the department. He keeps track of key performance indicators like operational efficiency, labour efficiency, and spoilage. Safety is also an important factor for the department manager and he is very keen on the safety measures within the department. The maintenance schedule is the responsibility of the department manager and he also makes sure that the company’s maintenance software is provided with input to keep it up to date. The maintenance function itself is very important to him and he is really the driving force behind the preventive maintenance program.

(32)

shifts per day, excluding the night shift, a team leader is present. Team leaders are also responsible for the output of production, safety, and quality. The team leaders report back to the department manager. The role of the team leaders in the maintenance function is limited to organisational issues. There is no influence on the maintenance scheme or process. Team leaders are involved for planning purposes as they determine the work schedules.

The process engineer is responsible for improving the production process. This is mostly done on project basis. For example, the process engineer is currently tracking and logging the condition of the tab. Severe recent quality issues demanded more stability in the process. Now every Monday the condition of the tab is checked and pictures of the tooling are logged. Purpose of this process is to see trends in deviation or other possible issues that arise over time. Although the process engineer is not directly involved in the maintenance function, projects like just mentioned can be valuable for the maintenance function which justifies the involvement of the process engineer in this research.

The technical assistant is responsible for all technical issues regarding the production lines within the department. In case of problems, the first lines of defence are the process technicians and if they cannot solve the issues it is passed through to the technical assistant. The technical assistant is, together with the process engineer, also involved in improvement projects of any kind. The maintenance function is coordinated by the technical assistant. He is responsible for coordinating the maintenance activities, ordering the required parts and supporting all employees involved in the maintenance function. In short, the technical assistant performs all maintenance related activities except the actual maintenance itself.

The process technicians are the technical specialists during each shift. Per shift there is always one process technician at work and he is the first line of defence in cast of any issues with the machines. When an operator cannot solve the issue it is passed through to the process technician. These technicians are also responsible for performing the actual maintenance on the machines.

(33)

4.1.1 Maintenance organization

Since one year, the maintenance manager focussed more and more on the preventive maintenance schedule. Twice per year an inspection takes place to identify components that need to be replaced during the following preventive maintenance period. Once per year the actual preventive maintenance (overhaul) is performed. On top of that the department uses maintenance software that supports the department manager with the schedule, and supports the maintenance personnel by showing which activities need to be performed. Next to the yearly maintenance schedule, additional activities such as lubrication or small part replacements are added to this system. When an activity is planned, the system makes a work order which is printed and provided to the process technician. The activities are carried out and confirmation is provided to the system.

4.1.2 Necessity of maintenance

(34)

Figure 14: Availability loss of production line

(35)

4.1.3 Results from interviews

This section discusses all findings from the interviews conducted with all the employees that are involved with maintenance. The goal of this section is to find all relevant factors that influence the maintenance decision in practice.

What factors can influence the maintenance decision?

To provide a clear picture of all the different factors, this section is structured according to the OEE model. Reason for this is that the OEE model clearly categorizes the different aspects that need to be considered in asset management. First the factors related to availability loss are discussed, after that the factors related to performance losses, and at last the quality losses are discussed.

Availability losses

According to the model of Jeong & Phillips (2001), availability losses consist of non-scheduled downtime, scheduled maintenance, unscheduled maintenance, R&D usage, engineering time, and setup & adjustment. In this section, non-scheduled downtime is referred to as the time needed for breakdown maintenance.

During the interviews it became clear that non-scheduled downtime is the most important issue at the case company. On a typical production day many short or long stops occur and each of these stops require a certain action. The amount of stops/failures is not only determined by breakdowns but also by line outages due to changes in product specification, safety checks, or machine errors. Where a large amount of failures can be solved by simple adjustments or machine resets, a great portion of failures requires corrective maintenance. Repairing broken equipment costs the company a lot of resources which restricts the company in the effort that can be put in improvement projects. The big issue with the amount of corrective maintenance activities is the influence of these failures on the quality of the product. Broken equipment often causes degradation of product quality around the time of failure but also hard quality failures. When these failures slip through all the safety checks in the process they form an immediate danger to the market.

(36)

are scheduled to determine need for maintenance. The issue with this preventive maintenance schedule is that it is based on gut-feeling and that the schedule sometimes is adjusted in case of high demand or lack of resources. The maintenance intervals are primarily based on experience and the inspections and maintenance activities are now used to determine whether these intervals are appropriate.

Unscheduled maintenance, R&D usage, and engineering time are not relevant for the production lines and are therefore left out. Setup & adjustment time is partly to do with performance losses and is discussed in the following section.

Performance losses

According to the model of Jeong & Philips (2001), performance losses consist of WIP starvation, idle without operator, other time, and speed loss. Also time lost with adjustments is discussed in the following section. Setup times are because of the dedicated lines negligible and therefore not relevant for this research.

At the case company, most performance losses are due to short failures which require machine resets, unblocking of machine, adjustments of tooling or equipment, recalibration of inspection equipment, or cleaning activities. The issue with the adjustments of tooling is that standardization is often lacking. Operators, process technicians or process engineers sometimes use different values for the same parts, or in other cases it is unknown what the actual settings should be, based on design specification. Recalibration of inspection equipment concerns the camera’s within the production line. The reliability of these systems is often point of discussion and a lot of resetting and recalibration is required to keep the production line going.

Quality losses

(37)

sampling procedure the amount of products that need to be checked is determined after which the actual inspection takes place. During the period of research multiple employees were scheduled for these activities on a daily basis costing a great amount of resources.

After identifying the issues related to the losses of the OEE model, this model is now used as a basis for a related model that shows all the possible interventions to an asset. These interventions are presented in figure 16.

Requirements

After identifying the different interventions at the case company, some other phenomena were noticed during the research. Most of these describe why the case company struggles with the maintenance function and why the company finds it hard to improve their maintenance plan. In a sense, these issues capture the requirements that should be in place to successfully implement a maintenance policy.

(38)

First of all, the availability of resources is an issue. When the focus is on improving the quality of the end product, the lack of resources is predominantly an issue for the development of the maintenance function. The high amount of line outages on a typical production day requires more than enough attention and little time is left for improvement projects. Operators are often too busy to fill in the downtime registration forms, the process technician has to be available to multiple lines during his shift, the technical assistant and the project engineer are involved in a lot of projects within the company and also need to be standby in case the process technician cannot solve the issues. This lack of capacity limits the development of the maintenance function. The focus is now mainly on getting the equipment up and running again instead of long term problem solving.

Another important factor is the budget available for maintenance. In practical applications, the maintenance function is, just as other functions within the organization, subject to a budget constraint. Within the case company the budget is not really a constraint on the daily maintenance activities but mainly on improvement processes regarding maintenance. Implementing changes or purchasing new parts can be a lengthy process as investments need to be approved by higher management. This causes irritation among employees. This irritation is strengthened by the feeling of the employees that their department does not get the same attention as the most recent, relatively new, department. They feel left behind and share the opinion that their department does not get the required funds for improving the process.

(39)

shorter than ten minutes it is just counted and in case of a failure longer than 10 minutes the actual time of breakdown has to be recorded. During the interviews and the meetings it became clear that operators do not always fill in this form correctly and that it regularly happens that at the end of the shift, the form is filled in as good as possible from memory. The data from this form is documented into spreadsheets but mainly for the registration of labour and not for downtime analysis. That this data is not used for downtime analysis has also another reason. The data that is recorded is most of the time not very useful in the sense that it can be used for improving the maintenance function. The downtime registration form only indicates the time that a certain machine has been down. The reason of outage is only in some cases described in a Microsoft word file logbook. Figure 15 shows that both sealmachines are responsible for most of the downtime. This machine however performs several steps in the production process but at which step the stoppage occurred is not clear. Whether the stoppage was caused by machine failure, set point shift, blockage or whatever other reason is unknown but hopefully an operator mentioned this in the logbook. Interviews showed that in most cases only real abnormalities are mentioned in the logbook. All this means that although all this data is gattered, it can be hardly used to say anything about machine wear and tear or specific parts/processes that have an influence on product quality.

As was mentioned in the literature background, the concept of imperfect maintenance is recognized in more and more researches. From the case study research it became clear that this is a very relevant issue in practice and can have a rather big influence on the production process. During the case research it was noticed that an issue sometimes reoccurs a couple of times shortly after a repair activity. This was not only the case after corrective repair actions but also after extensive preventive maintenance activities. The issue here however is to find out what the real issue is. Are the repairs performed incorrectly, are repaired or overhauled parts appropriate for the job, or lies the problem with the settings used after the maintenance action. More effort should be put in on finding the real root cause of the problem. On top of that the maintenance activities are performed by different people with different experiences and knowledge. It is therefore hard to say that each person performs the repair in exactly the same manner as the others.

(40)

separately. During the case research it happened regularly that machine settings were a point of discussion. It happened that different process technicians used different settings for the same part which caused functional problems with transportation. On another occasion it appeared that due to cleaning and maintenance activities the stamp in one of the sealmachines gradually reduced in length. Although the margins are very small, this will eventually cause problems when the loss of material is not compensated for in the machine settings. The general issue here is standardization. During the interviews it became clear that the goal is to achieve standardization as much as possible but at this moment this is far away. However, it has attention from management and a lot of effort is directed towards this standardization. Another interesting result from the interviews that showed the importance of setting management relates to the speed of the production line. Recently the speed of the line was increased to be able to meet market demands, but this increase also causes some issues. Not only are the machines subject to increased wear and tear, it also causes issues with key parts for quality. An example of this is the process in the sealmachine. When the speed is increased it becomes harder to properly feed material through the machine which can cause key quality issues like leaks or tabs on the product that are off position. Increasing the speed of another part like the compound injector might cause gaps in the compound. These gaps cause leak cans and leak cans is the number one quality issue. So, standardization off and proper research on the settings of machines is absolute key when the quality of the end product is a priority.

Discussion

(41)

interviewees stated that eventually this could be very valuable information when the equipment functions well, but that at this time the priority lies at structured planned maintenance and keeping the equipment running.

4.2 Design and development.

Dekker (1996) stated in his research that maintenance is a generic term that covers a wide variety of factors which means that there is no general model that covers all possible causes. He also states that there are a multitude of models available but that there is a lack of knowledge on which model should be applied to a certain practical situation. Therefore the focus should be on unifying existing models instead of publishing new ones. This is exactly what this research is trying to do. As mentioned before, this research aims at providing a decision model that helps the company with selecting the appropriate maintenance policy. Therefore existing models are reviewed and an attempt is made to integrate these models with knowledge obtained at the case company.

Before the final model is discussed, the data from the literature review and the case study are integrated and discussed to find out which concepts the model should cover.

4.2.1 Use of multiple maintenance policies

(42)

4.2.2 Factors that influence the maintenance decision.

In section 4.1.3 several different factors were mentioned that influence the maintenance decision. Some of these factors have a direct influence on the maintenance decision while others are more organizational requirements that should be in place for setting up an appropriate policy. Table 1 shows the factors found through the literature review and the case research. When looking at the importance of the different factors it became clear that acquiring the appropriate data is the most important when the goal of a maintenance policy is to improve the quality of the end product. When there is no information available on what parts experience the most wear or what parts cause the most quality issues it becomes impossible to design an appropriate policy. After that the standardization of settings is important to make sure that all parts and systems run according to the right settings and that all maintenance staff is aware of these settings. The availability of resources should be known before designing a maintenance policy as it drives or limits the possibilities for the maintenance function.

(43)

Factor Maintenance decision Machine condition Perfect/Imperfect maintenance x Setting management x Data requirements x Inspection/sampling schedule x Maintenance method x Human interference x Availability of resources x Complexity of equipment x

Table 1: Factors influencing the maintenance decision

The factors mentioned in this table serve as input for the model proposed in the following section.

4.2.3 Design solution

Based on the knowledge acquired through the literature review and the case study a rather extensive model is proposed that indicates which factors need to be taken into account when designing a maintenance policy. Not necessarily all factors have to be determined as for some situations a certain factor will not have any influence. These factors are however intentionally incorporated as the goal of the model is to build a generic design. Not taking a factor into consideration is always better than leaving out a factor that could have an important influence. The model is roughly constructed according to the proposed framework of Wayenbergh & Pintelon, (2002), presented in figure 5 Figure 18 shows the proposed model. How this model works is explained next.

(44)

setting management within the company. Settings must be standardized and equipment must be able to run properly under those settings.

In the second step, a decision tree is used as the one proposed by Waeyenbergh & Pintelon (2002). For every part the decision must be made whether it is critical for the quality of the end product or not. If the answer is yes, condition-based maintenance is the most appropriate strategy as set-point shifts can be noticed at an early stage in the process. If the part is not critical for quality, it must be determined whether or not it is critical for the system. As the performance of the system is always important, it is not realistic to solely base the maintenance decision on the criticality to quality. When the part is not critical for the system it is best to subject this part to corrective maintenance. When the part is critical to the system it must be decided whether time-based maintenance or condition-based maintenance is most appropriate.

(45)

(46)

5. Discussion

In practice it is impossible to focus on a single maintenance policy and therefore it is important to set up the maintenance function as a combination of all maintenance policies. In this research an attempt is made to identify the different factors that influence the maintenance decision and incorporate these in an integrated model following the maintenance concept development model of Waeyenbergh & Pintelon, (2002).

The goal was to find an answer on the following question:

The selection of a maintenance policy follows in general the steps proposed in the model by waeyenbergh & Pintelon, (2002) presented in figure 5. This means that before a maintenance policy can be implemented, the objectives need to be clear and the appropriate resources need to be available. From the case research it became clear that although the objective within the company is to improve the quality of the end products, the company stays behind in the organisation of the maintenance function. Large amounts of resources are spent on repairing equipment, data gathering, and performing preventive maintenance but the results of these efforts stay behind. This research shows the requirements for the successful implementation of a maintenance policy.

(47)

Contribution

The main contribution of this thesis is that it provides a framework to companies, that depend on high quality products, how they could organize their maintenance function. By making product quality a number one priority instead of using it as a part of a metric like the overall equipment effectiveness, it becomes easier to spot problem areas within systems and to intervene in the process more quickly. To do so a model is proposed that proposes the most appropriate maintenance policy for parts or sub systems within the larger production system.

Recommendations

The case company should improve and standardize its data gathering procedure and should carefully analyse how the data is used in their attempts to improve the process. In the current situation a lot is effort is lost because the all the data gathering takes a lot of time but the actual interventions that come through this data is very limited.

More effort should be put into identifying parts that have a key influence on the quality of the end product. Although, based on experience, the company can identify the most problematic parts or sub systems, it is impossible for the company to pin point the exact reasons of failure. By better monitoring sub systems, the most critical parts can be identified and appropriate measures can be taken.

The company should discuss the opportunities for implementing condition-based maintenance on key parts or sub systems within the process. Implementing such a policy provides the company with an opportunity to intervene in the process whenever a set point shift or a quality shift is noticed. This prevents a loss of resources caused by producing additional products and rework.

Limitations

The biggest limitation of this research is that it is focussed at improving the quality of the end product. Companies who are more interested in a cost effective maintenance function might be better off by setting a different objective as a starting point for the maintenance function.

(48)

6. References

1) Ab-Samat, H., & Kamaruddin, S. (2014). Opportunistic maintenance (OM) as a new advancement in maintenance approaches. Journal of Quality in Maintenance Engineering,

20(2), 98–121.

2) Ahmad, R., & Kamaruddin, S. (2012). An overview of time-based and condition-based maintenance in industrial application. Computers & Industrial Engineering, 63(1), 135–149. 3) Alaswad, S., & Xiang, Y. (2017). A review on condition-based maintenance optimization

models for stochastically deteriorating system. Reliability Engineering & System Safety, 157, 54–63.

4) Alsyouf, I. (2009). Maintenance practices in Swedish industries: Survey results. International

Journal of Production Economics, 121(1), 212–223.

5) Arabian-Hoseynabadi, H., Oraee, H., & Tavner, P. J. (2010). Failure modes and effects analysis (FMEA) for wind turbines. International Journal of Electrical Power & Energy Systems, 32(7), 817–824.

6) Arunraj, N. S., & Maiti, J. (2007). Risk-based maintenance—Techniques and applications.

Journal of Hazardous Materials, 142(3), 653–661.

7) Asadzadeh, S. M., & Azadeh, A. (2014). An integrated systemic model for optimization of condition-based maintenance with human error. Reliability Engineering & System Safety, 124, 117–131.

8) Azadeh, A., Sheikhalishahi, M., Mortazavi, S., & Jooghi, E. A. (2016). Joint quality control and preventive maintenance strategy: A unique taguchi approach. International Journal of

System Assurance Engineering and Management.

9) Barlow, R., & Hunter, L. (1960). Optimum preventive maintenance policies. Operations

Research, 8(1), 90–100.

10) Ben‐Daya, M., & Duffuaa, S. O. (1995). Maintenance and quality: The missing link. Journal

of Quality in Maintenance Engineering, 1(1), 20–26.

11) Bousdekis, A., Magoutas, B., Apostolou, D., & Mentzas, G. (2015). A proactive decision making framework for condition-based maintenance. Industrial Management & Data Systems,

Designing a practical asset intervention policy based on product quality degradation