Abstract
Purpose
To prevent the assets from failing it is of utmost importance that mitigating actions are taken in time and that these are economic sound decisions. In general, physical assets degrade over time and thus investments need to be done to keep the asset’s condition according to standard. Asset managers are responsible for the assets and have the tough job of making a replacement decision. This decision needs to be made at the end of its Remaining Useful Life (RUL). Both for the purpose of extending the literature and for adding to the knowledge of practitioners that are responsible for managing the asset, it would be beneficial to gain more insight into the decision-making process of selecting an asset investment. Moreover, many companies also face the situation where their assets are reaching or are even beyond their design life. Furthermore, many asset decisions at other companies are also made based on merely technical expertise and gut-feeling. 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 asset investment decision, given a certain RUL assessment?
Method
In this research a design science methodology is used to design an artefact that complies with all the design criteria. Data will be gathered through a literature review and at a case company: AkzoNobel Delfzijl. The AkzoNobel case study gives more insight in the way a production company has organised the asset investment decision. The results from this case study can be used in a more generic way, as the design solution is applicable at other capital-intensive companies.
Findings
3
Preface
With this thesis I conclude my masters programme Technology & Operations Management at the University of Groningen. The goal of this thesis is to expand the current knowledge on asset management. Insights on the asset investment decision process were obtained by doing a case study at the salt business unit of AkzoNobel Delfzijl.
This practically oriented research was a true informative and challenging experience of which I have learned a lot. Therefore I would like to thank AkzoNobel Delfzijl for this research opportunity and a special thanks goes out to Piet Drijfhout, Henk Assink and their colleagues, who were very helpful and provided me with all the needed data.
Moreover, I would like to thank my supervisor dr. Jasper Veldman who challenged me during my BSc thesis and now once again to achieve my best effort in this research. His constructive feedback has helped me to specify my research project and follow the right steps in order to achieve a satisfying result. Furthermore, I would like to thank my co-supervisor dr. ir. Wilfred Alsem who provided me with valuable feedback and organised the peer-review sessions. Finally, I would like to thank my family and friends for supporting me and for their encouraging words throughout this challenging period.
Groningen, June 2016
1. Introduction
In the beginning of the second half of the 20th century, the economy was booming and thus the demand for physical equipment installations was high. Only, their design life was most of the times only about thirty years but most of these assets are still operating. This is also the case for many process industry assets. Moreover, a company’s assets are mainly responsible for generating the revenue. To prevent the assets from failing it is of utmost importance that mitigating actions are taken in time and that these are economic sound decisions.
Ruitenburg, Braaksma and Van Dongen et al. (2014) argue that a thorough understanding of the asset’s complete lifetime is of key importance to grasp the full potential of the asset. In general, physical assets degrade over time and thus investments need to be done to keep the asset’s condition according to the standard. Asset managers are responsible for the assets and have the tough job of making a replacement decision. This decision needs to be made at the end of its Remaining Useful Life (RUL).
In recent articles much has been written about the RUL assessment and the development of prognostic models that are used to predict the RUL of an engineering asset (Sikorska et al., 2011). Estimating the RUL of an asset is highly important in condition based maintenance (CBM) (Wang & Zhang, 2008) and influences the planning of maintenance activities, spare parts provision, operational performances and profitability (Papakostas et al., 2010). Moreover, the asset replacement decision and its many aspects have been touched upon. Single and multiple asset replacement models are reviewed where also the asset’s age and utilisation have been taken into account (Bethyne, 1998; Hartman, 2001). Also the replacement policy of a fleet of assets has been researched (Figliozzi et al., 2011).
However, in literature not much attention has been given to the exact steps that need to be taken to come to an asset investment decision. Most articles only focus on either the merely technical aspect of an asset and assess the RUL or, researchers analyse the mathematical models that support certain replacement decisions. Still most asset investments are made based on a gut feeling (Bouyssou et al., 2006) as asset manager do not know what the best asset investment decision is (Jones & Sharp, 2007)
Moreover, the assets are highly valuable to the company hence a large amount of money is involved. Also, asset failures near its remaining useful life tend to be more costly than preventive refurbishment or replacement as these failure occur at inconvenient times, require service restoration action and temporary measures (Wijnia et al., 2006). Therefore, it is important to have a systematic process of coming to a well-grounded investment decision.
7 The large and wide variety of RUL models and asset replacement models make it difficult to quickly establish the best asset investment decision-making process. Merely the use of these models would result in a shortcoming of the investment decision-making process, as the RUL and replacement models do not encompass all the necessary factors that play a role in the investment decision-making process. Especially as the investment decision comes with great responsibility for the decision maker and is required to be well funded to get accepted by higher management (Hastings, 2015). Moreover, integrating the RUL and replacement model is not trivial because there are uncertainties in the RUL estimate and there is an overreliance on technical and operational justifications (Woodhouse, 2011 ).
Both for the purpose of extending the literature and for adding to the knowledge of practitioners that are responsible for managing the asset, it would be beneficial to gain more insight into the decision-making process of selecting an asset investment. The asset investment decision-making process in this research is defined as all the steps that have to be taken to come to the final asset investment decision. 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 asset investment decision, given a certain RUL assessment?
In this research a design science methodology is used to design an artefact that complies with all the design criteria. Data will be gathered through a literature review and at a case company: AkzoNobel Delfzijl. The AkzoNobel case study gives more insight in the way a production company has organised the asset investment decision. The results from this case study can be used in a more generic way, as the design solution is applicable at other capital-intensive companies. Moreover, many companies also face the situation where their assets are reaching or are even beyond their design life. Furthermore, many asset decisions at other companies are also made based on merely technical expertise and gut-feeling (Bouyssou et al., 2006).
This paper is structured as follows. First a literature background will give further information about the topic. Then more details about the methodology are discussed. After this the problem investigation is explained including more details about the case company. The design and development chapter follows next, after which the design is evaluated. This research is finished up by the chapters covering the discussion and conclusion.
2. Literature background
In this chapter background information is given on the research subject.
First basic information is presented about the asset life cycle of an asset. At the end of this asset life cycle the asset has approached its end-of-life. The time till the end-of-life can be determined by estimating the remaining useful life. The different aspects of the term ‘useful’ are discussed in the Lifetime Impact Identification Analysis of Ruitenburg et al. (2014). Afterwards, the technical RUL models and economical RUL models are discussed. After the RUL determination a replacement decision needs to be made and therefore first an asset replacement model is presented and afterwards the replacement decision options.
2.1 Asset lifecycle
Currently companies are facing challenges to maintain and improve the service levels of their assets although only having access to limited funds. Asset management is therefore increasingly important to find a solution for this issue (Allbee, 2005). Asset management is defined (Schuman & Brent, 2005) as “a strategic, integrated set of comprehensive processes to gain greatest lifetime effectiveness, utilisation and return from physical assets” (p. 567). To increase the value of the assets, the asset management needs to encompass the following phases: conceptual design, preliminary, detail design and development, production and/or construction, utilisation and support and retirement and disposal. In figure 1 these phases are depicted showing a complete life cycle. Figure 1 – Asset lifecycle (Blanchard & Fabrycky, 1998) An asset enters the last stage of the life cycle, the retirement and disposal stage, when for example machine tools are (completely) worn out, a truck has broken down or electronic equipment became obsolete. In other words the asset has attained its technical useful life. Another option is when the asset is no longer economically valuable to repair or maintain. For instance the physical deterioration causes the production capacity to decline, which harms the economical value of the asset (Hulten & Wykoff, 1996). Mostly the economical remaining useful life is earlier than the technical remaining useful life.
9 environmental, social and technical factors and performances” (p.286). This is thus another model and view on the asset lifecycle than the one portrayed by Blanchard and Fabrycky.
Figure 2 – Asset lifecycle management model (Haffejee & Brent, 2008)
First, there are two drivers of change, internal and external. The internal drivers of change are defined as those factors that are driven from within the organisation and that cause changes within the same organisation. The external drivers of change are defined as those factors that emanate from outside the organisation, but cause changes within the organisation. The assessment of the strategic assets is done by taking into account the economic or financial impacts (life-cycle costing or total cost of ownership), the environmental impact(s) (environmental incursions in terms of chemical spillages, land refills, etc.), the social impact(s) (disasters that affect people in anyway), the technical impacts (efficiency and effectiveness) and the technical losses. The asset renewal decision refers to a decision that needs to be made regarding the future of the asset. A sound business case, unless there are safety and/or legislative reasons for asset renewal, needs to be compiled to support the asset renewal decision. The strategic assets are operated and maintained in accordance with predetermined guidelines, standards and specifications to achieve the design life with minimal costs, social and environmental impacts. The total performance of the asset is measured and managed in accordance to certain specifications with respect to economic, environmental, social and technical performances. The recorded data of the performance is collected and verified against design performance parameters. Lastly a detailed asset register is made and updated frequently to reflect the performance of the assets.
In respect to the RUL estimation of an asset the asset lifecycle model ‘measures’ the RUL constantly as the steps are depicted in a loop. In the model the storage of data is included on which a trend can be established to indicate if the asset performance deteriorated to an unacceptable condition. The recorded data is stored and with this data an analysis can be done to estimate the RUL. Haffejee and Brent (2008) do include the RUL assessment although do not directly name it this way.
2.2 Lifetime Impact Identification Analysis
Ruitenburg et al. (2014) argue that a thorough understanding of the asset’s complete lifetime is of key importance to grasp the full potential of the asset. There could be problems with data collection and the quality of data, which limit the availability of the ALCM and its potential to deliver a deep understanding of the asset. Therefore the authors have come up with lifetime impacts, which are trends or events that “may have a positive or negative influence on the remaining useful life of the asset” (p. 204). Five main perspectives on the remaining lifetime of assets were identified: technological, economical, compliancy, commercial and organisational (TECCO). The technological perspective looks into the technical specification and the amount to which the asset can comply with these specifications. The economical perspective covers the cost of operating and maintaining the asset and this perspective is therefore closely related to the Life Cycle Costing literature. The compliance perspective looks at a wide range of different aspects of the ‘licence to operate’ of the company. Examples of these aspects are sustainability, safety, working conditions and sectorial norms. The commercial perspective has to do with the compliance to market demand of the asset and its products. In this perspective an analysis is made on the asset’s output in relation to the output attractiveness to the customer. Lastly, the organisational perspective looks at the ability and willingness of a company to operate the asset. Specific knowledge might not be present or there is a limiting ability to maintain the asset in the desired condition.
11 product that the asset produces and the knowledge and expertise necessary to operate the asset respectively.
As now the term RUL is more specified, further details will be given on how the RUL can be estimated. In the following section the theory about what aspects need to be taken into account when a replacement needs to be made will be specified. First an overview is given of the technical RUL models and afterwards the economical RUL models will be discussed. Lastly, an overview of the literature background will be presented.
2.3 Technological RUL models
‘Technology’ or ‘technological’ can be interpreted in multiple meanings. In this study the technological perspective is seen as the physical state or condition of an asset, which can be measured by for instance vibration measurements or wall thickness. With the technical condition also the social and environmental condition of an asset are meant, as for instance human safety or ecological impacts. The technological useful life an asset can be defined as the period over which the asset is expected to last physically, to when replacement or a major overhaul is needed. In other words, an asset starts to operate from its brand-new condition to a status in which it can no longer be used in the normal operating state and must be retired (Woodward, 1997). Degradation can be recognised as the deviation in performance from that when the asset was new (Li & Nilkitsaranont, 2009). In figure 3 different degradation patterns are depicted where the degradation graph crossed a certain threshold ranging from the pessimistic RUL through the predicted RUL towards the optimistic RUL.
Figure 3 – Degradation patterns resulting in different RULs according to a certain failure pattern (Li & Nilkitsaranont (2009)
In the paper of Xiongzi et al. (2011) the authors stated that the estimation of the RUL is naturally dependent on the probability density function (PDF) and this influence is clearly depicted in figure 4. The estimate of the RUL can be seen as a continuous random variable as the options are unlimited. For a continuous random variable the probability that X takes a specific value x is zero therefore the probability that X falls within a range is used. This calculation is done with a probability density function and can be written as P(a < X < b) (The Pennsylvania State University, 2016).
Figure 4 – Degradation graph indicating the RUL of an asset based on a predicted trajectory (Xiongzi et al., 2011)
In the last years much has been written about the RUL and especially about RUL models. According to the review paper of Ahmadzadeh and Lundberg (2014) RUL models can be divided into four categories, namely: physical, experimental, data-driven and hybrid. The four categories are reviewed in table 1 to give an overview of the description, input requirements, the purpose and several examples.
After the technological RUL models the main economic RUL models are presented.
2.4 Economical RUL models
The economical remaining useful life, which is a game between revenue and cost (Wang et al., 2015), can be modelled through increasing operating and maintenance expenses (OpEx) and/or a decrease in salvage value of the asset. The economic life of an asset is “the period until economic obsolescence requires replacement with a lower cost alternative” (Gorjian et al., 2009). In other words, the period until which an asset is no longer of value economically, however it still could be technologically above the set limit. The economical RUL of a piece of equipment set thus the optimal replacement age that results in the minimal acquisition, operating, maintenance and salvage costs. The following models take these costs into account: life-cycle costing and total cost of ownership. Moreover, this sub chapter will be concluded with an explanation about the Capital Expenses (CapEx) and Operating Expenses (OpEx).
2.6.1 Life-cycle costing
Life cycle costing (LCC) can be defines as “the total cost of the system or product under study over its complete life cycle or the duration of the period of study, whichever is the shorter” (Norman, 1990). Life cycle costing can be used as a forecasting tool to weigh alternative capital investments with the aim of ensuring the optimum value from physical assets. The method comes down to expressing all future costs and benefits in present day values (Taylor, 1981). LCC was originally designed for purchasing purposes in the US Department of Defence and is used most often in the construction industry (Woodward, 1997).
Taylor (1981) defined eight types of costs that need to be taken into account when assessing the total life-cycle cost of an asset: specification, design,
acquisition/manufacture/build, installation, commissioning, operating,
maintenance and disposal. The author developed a model to structurally include all the life-cycle costs, see figure 5. First the requirements of the assets are identified. In this step environment and objectives of the enterprise over a period of time are considered. In the second step the selection of physical assets is made which match the needs at the lowest costs over the life of the asset. After this step monitoring of the actual costs will take place. The planning activity is made up of forecasting the costs and also of understanding the relationships among the different costs at every phase of the life-cycle.
Figure 5 – Life-cycle costing model (Taylor, 1981)
Woodward (1997) developed a model that monitors the asset throughout it entire life cycle from the development stage to the disposal stage, see figure 6. The model can be used to maximise the value for money in the ownership of the asset by taking into account all the cost factors with respect to the asset during its operational life. The key is the optimisation of the trade-off between the costs factors which in the end will result in the minimum life-cycle cost of the asset. The investment consideration can only be made by estimating the costs on a whole life basis before procuring the asset. Woodward (1997) divided the LCC model into eight steps: establishment of operation profiles, establishment of utilisation factors, identification of all the cost elements, determination of the critical cost parameters, calculation of all costs at current prices, escalation of current costs at assumed inflation rates, discounting of all costs to the base period, summing up discounted cost to establish the net present value.
Figure 6 – Asset lifecycle costing model (Woodward, 1997)
2.6.2 Total Cost of Ownership
17 implemented the TCO analysis and the author has developed a framework for the TCO implementation. Moreover, Ellram (1994) also discussed a taxonomy to classify TCO models with respect to the different types of buy. Bennett (1996) has presented a TCO model that was used by Compumotor, which is a manufacture of flow control equipment where this company has linked the TCO to its activity-based costing system. This provided an integrated approach to the TCO analysis.
2.6.3 Equivalent Annual Cost model
Figure 7 – CapEx/OpEx as function of the asset lifecycle (Woodhouse, 2011)
After the RUL assessment has been completed the asset manager can think about what kind of investment decision has to be made. There are multiple options available to extend the RUL of an asset. These options are discussed below. 2.5 Asset investment decisions
In the article of Gitman (2003) the author stated, “All capital budgeting decisions can be viewed as replacement decisions. Expansion decisions are merely replacement decisions in which all cash flows from the old asset are zero” (p. 362). Terborgh (1958) distinguished between the old, current asset with which the company is producing its products and the (possible) new asset that will be replacing the old one. The current asset is called the ‘defender’ and the new asset is called the ‘challenger’. 2.5.1 Investment options A company has many options to have the asset up to the required reliability and conditional standard. Certain considerations can be made if a company registers that the condition of an asset is decreasing. First of all, the company can decide to leave the asset as it is and let it run to failure. Moreover, it could also refrain from an (big) investment but keep maintenance policy in place and postpone the replacement decision. The replacement decision could be further specified by deciding to only replace one part or multiple parts, doing an overhaul or replacing the entire asset all at once (Haffejee and Brent, 2008). As can be seen in figure 8 of Li et al. (2006), it is possible to invest multiple times in an asset to extend the RUL. In this figure only maintenance activities are shown, larger replacement action will likely result in a larger increase towards the original condition of the asset as can be seen by the arrows in figure 9. In the end the maintenance actions come down to buying time, i.e. extending the point of end-of-life and thereby extending the RUL.
19 Figure 8 Relationship between the value, time and PM for an aging asset (Li et al., 2006) Figure 9 – Reliability as function of time with repair and replacement decisions made to increase the reliability (Schneider et al., 2006)
Figure 10 – Assessing index system of transformer economic life (Wang et al., 2015)
21 2.6 Conclusion
3. Methodology
A real world improvement problem is best studied by making use of the design science methodology (Karlsson, 2010). This methodology focuses on real world organisational problems through the use of scientific knowledge and local facts (van Aken et al., 2012) to create innovative artefacts that are the solution to the problem (Hevner & Chatterjee, 2010). Not only explicit but also tacit knowledge is used in the design science methodology, which is developed through experimental learning in a creative process of reflection-on-action (Van Aken & Romme, 2012). The goal of this paper is to answer the following question: How can a decision-making model be designed that supports the selection of the right asset investment decision, given a certain RUL assessment? According to many authors (e.g. Eekels & Roozenburg, 1991; Nunamaker et al., 1991; Hevner et al., 2004) the following design science steps need to be taken for a solid research execution: Problem identification, make solution objectives, design & development of artefact and evaluation. In figure 17 these steps are depicted graphically. Below, each step will be discussed in more detail. The artefact will be composed on a ‘function’-level. This means that the unit of analysis is an entire asset instead of for instance a part of equipment. This way the final model can be used for other assets within AkzoNobel and possibly other companies. To make sure the model proposed delivers what it set out to do, the model will be validated by using it to analyse an air pressure pump. This asset has been chosen for three reasons. First, the data monitored for this asset is extensive which allows for a good historical data analysis. Secondly, the asset has a clear and definable function, which aids gathering the data necessary for the analysis. Thirdly, the data indicate that this asset already should have been replaced which allows for a hindsight-analysis to see what action should have been taken to avoid the decrease in condition level and what the possible consequences would have been.
Figure 13 – Design science methodology steps
Problem
23 3.1 Problem identification
The first step is to get all the information about the problem and justify the value of a solution. The problem solver needs to know what the problem exactly is. This way it can make an assessment whether the proposed design will be the best solution to the problem and can predict what would happen if the design were to be implemented (Wieringa, 2010). It is suggested that the problem needs to be described conceptually so that the solution can capture the complexity of it (Peffers et al., 2007).
In this research the problem identification is addressed by sub question 1.
Sub question 1: What is the managerial problem of AkzoNobel Delfzijl?
3.2 Objectives of the solution
It is imperative for designing an artefact to have certain objectives in place to evaluate and rate the artefact on. The objectives that need to be set can be deduced from the problem definition and from the design science literature (Peffers et al., 2007). The problem definition related objectives will be retrieved through a stakeholder analysis. The other objectives are established through a literature review on design science research methodology. The objectives have to be specific, otherwise it is not possible to assess afterwards if the goal has been attained or not. The issues addressed above are divided into sub question 2, 3 and 4. Sub question 2: Who are the stakeholders? Sub question 3: What are the goals of all the stakeholders? Sub question 4: What are the design science research criteria? 3.3 Design & development
Landry et al. (1983) came up with three types of validation: conceptual validation, logical validation and experimental validation. Conceptual validation requires an iterative process of going from the problem situation to the conceptual model until the acceptable level of validity is achieved. Logical validity is concerned with capacity of the formal model to correctly describe the problem situation as is defined in the conceptual model. Lastly, experimental validity refers to the quality and efficiency of the design mechanism. Mainly it is concerned with the type of solution, the efficiency of acquiring the solution and the sensitivity of the solution to changes in parameters of the model. As was stated before, the model will not be implemented and therefore in this section only the main issues will be addressed on how a correct implementation might look like. Radnor et al. (1970) came up with three ways of looking at the implementation of a model. The first viewpoint considers the implementation as “a transition process that takes place between successive stages in a work-flow pattern”(p. 968). The second viewpoint is concerned with the implementation being a special case change or adaption to the organisation. Lastly, implementation can be seen as a continuous process along all phases of a project instead of only looking at the improvement as the last stage in the design process. The main issue therefore is to always have the implementation phase in the back of the head when going through the earlier phases. In other words, it can be seen as a design-for-implementation tactic. The solution evaluation phase and issues at implementation will be covered by sub question 7 and 8:
Sub question 7: What is the validity of the proposed investment decision-making artefact?
Sub question 8: What will be the main issues when the artefact will be implemented in the salt business unit at AkzoNobel Delfzijl?
3.5 Data collection
The data for this study is collected in multiple ways. Firstly, the general theory is gathered through a literature review and is discussed in the literature background. To get a good sense of the problem and its context, interviews are held with multiple employees in the asset management and maintenance department of AkzoNobel. Moreover, the data concerning the validation and implementation will be gathered through hosting expert group meetings where the proposed design will be discussed and thoroughly analysed on its validity and possibility of implementation.
25
4. Problem investigation
In this chapter the problem of the case company will be discussed. Before analysing the specific problem, more information is given about the company itself and the production process. To come up with a solution for this problem, several matters need to be explored and investigated. The first item is the analysis of the different stakeholders that play a role in the investment decision-making process of an asset. The second item is concerned with the criteria of the stakeholders for the decision-making model. This chapter will conclude with a list of design criteria on which the model will be assessed. The next chapters will cover the design and development, evaluation of the designed model, the discussion and the conclusion.
4.1 Case company
This research has been done at AkzoNobel and therefore a small introduction will be made to give background information on the company itself. Furthermore, the production steps of the salt process are presented.
4.1.1 AkzoNobel Delfzijl
AkzoNobel makes a wide variety of products in many different factories and plants around the globe. The plant of AkzoNobel in Delfzijl is part of the business category Industrial Chemicals and is located at the “Chemiepark Delfzijl” with several other production companies. AkzoNobel Delfzijl has four business units and one joint venture. The four business units are: the salt factory, chloroacetic acid factory (MCA), membrane electrolysis factory (MEB) and Delesto and this factory produces electricity and steam. The Delamine plant is a joint venture between AkzoNobel and the Japanese Tosoh Corporation and produces ethylene amines. The research of this thesis will purely focus on the AkzoNobel salt factory. As salt is one of the most important raw materials that is used by other factories at the site, the salt factory is highly important within the production chain at the Chemiepark Delfzijl. Each year the salt factory produces around 2,7 million ton of salt.
4.1.2 Production steps
Salt dissolved
into brine Purikication of brine
Evaporation of water within brine Extracting remaining water by spin-drier 4.2 Management problem The assets on site are between 10 year and 50 years old. This difference in age is due to the fact that a relatively new plant, Salt factory D, has been constructed about ten years ago. For the older assets, 30 years and over, it is necessary to investigate how much time they still have left to be used effectively and efficiently. Especially their Remaining Useful Life has to do with the amount of operating expenditures (OpEx), like regular maintenance and capital expenditures (CapEx), which AkzoNobel has set on large asset investments of €50.000 and more. Currently the assets are assessed according to the knowledge and expertise of the employees (“hunch”) and according to a risk matrix on the following items: safety, environment effect, cost effect and loss of capacity. If the impact of a failure is low and the chance of this happening is low as well it will be indicated that the risk is green, which means that no immediate action has to be taken. However, if either the impact is severe or the likelihood of an incident is high, it will be marked as a red risk and means corrective maintenance should be undertaken. It goes without saying that a risk that is highly likely to happen and in the same time has a severe impact will be marked as a red risk as well. After the assessment of the assets, the red indicated maintenance projects will be used in a long term asset planning (LTAP). The main issue at hand here is that the investment decision is made based on personal expertise and opinions. There is currently no systematic way of assessing the RUL of an asset and especially not a way to assess the consequences of certain maintenance or replacement strategies. As a result, projects are postponed year after year until it (suddenly) becomes for instance too hazardous to operate and the asset’s functionality is limited or worse, the operation with this asset needs to cease until maintenance has been done. As AkzoNobel is a big multinational, the process of requesting maintenance/project funds takes at least a year and therefore bigger projects can not be planned and executed within a month when an asset (suddenly) has reached its end-of-life. Therefore a systematic method of assessing a physical asset within the Salt factory is highly useful for the asset practitioners. This way they can anticipate on the deteriorating condition of the assets and plan the maintenance or renewal activities accordingly.
4.3 Stakeholders
In this section the stakeholders of the proposed problem will be established. As this research has been initiated by the asset managers (asset group) of the salt business unit of Delfzijl, they can be classified as the main stakeholders. This asset management group consists of maintenance management, maintenance support and reliability engineers. The asset manager has the main responsibility for this department and is in charge of coordinating and managing all the assets
27 of all four factories (Salt, Delesto, MEB and MCA). The input for the RUL assessment comes from the maintenance manager and reliability engineer and they are therefore key figures in the assessment of the asset. The maintenance manager is responsible for the condition of the assets, which comes down to keeping track of the preventive maintenance program. As the maintenance manager is also concerned with the production process, this department measures and assesses the asset condition as well. For an overhaul or large replacement the maintenance manager delivers an inspection report, which is sent to the reliability engineer. In a monthly meeting the asset group, the technology manager, the plant manager and the production manager discuss the current state of the assets after which they take measures accordingly. The reliability engineers have to guard the reliability of the assets. Their main task is to design solutions for existing issues that are not executed by the maintenance department because they have at that moment a limited amount of (human) resources or not the specific expertise to resolve the problem. Mainly in large projects the reliability engineers are involved.
The technology manager is responsible for the development and condition of the process, all that what happens within the installations, like the warmth diffusion, crystallisation processes or the influences of the process on the construction of the plant like corrosion. Besides this, the technology department takes care of the product quality and process safety. Also the improvement of the process efficiency is the task of the technology department in which they mainly focus on energy efficiency and decreasing the usage of chemicals. Within the RUL assessment process, the technology department comes into play when finding a solution to extend the RUL of an asset is concerned. Possibly, one or multiple process variables could be adjusted so that the asset’s RUL could be prolonged.
The plant manager is in charge of the entire factory and is mainly focused on the strategy and how the trends of the salt market and the movements of the energy market influences this strategy. The plant manager stated that one of the important issues is to also take into account the current market situation. If the plant has to produce on 100% of its capacity due to high demand or only use 70% of the capacity as the demand has fallen, it will affect the way the assets are managed and therefore what the final assessment of the asset will be. The model proposed in this research only is applicable to those assets that will be used for the upcoming 10 years and therefore the plant manager is not a stakeholder for this model. In the MSc thesis of Michiel Zantinge, the strategy aspect is discussed in the systematic overview to come to a LTAP.
of an asset (group) is an issue and cuts back the RUL significantly, the production department can find a solution to this issue and thereby extend the RUL.
29 4.5 Design criteria
Designing an artefact is not a one-time action but an iterative process. The paper of Hevner et al. (2004) has been cited by many authors and in their paper a conceptual framework with design science research guidelines is presented that aids in the understanding, execution and evaluation of design science research. The first guideline is to design as an artefact, which means that the research needs to produce a viable artefact in the form of a construct, method or model. The second guideline is concerned with the problem relevance and covers the goal of the research and whether artefact made is a technology-based solution to an important and relevant business problem. Thirdly, Hevner et al. (2004) identified that the utility, quality and efficacy of the artefact must be rigorously demonstrated through well-executed assessment methods. Fourthly, the research needs to rely upon application of rigorous methods in both the construction and evaluation of the design artefact. Lastly, there needs to be a contribution to literature in the areas of the design artefact, which is in this case the asset management literature. The contribution to literature can be assessed by analysing the novelty, significance and generalizability of the design solution (Hevner et al. 2004). In table 4 the list with design criteria is listed. The stakeholder goals can be seen as additions to the design guidelines. Mainly the problem reference guideline and design assessment is substantiated by the goals of the stakeholders. Moreover, the reliability engineer stated that the model needs to be generalizable and this goal is listed under the research contribution.
Design Criteria Description Design as an
Artefact The research needs to produce a viable artefact in the form of a construct, a model or a method Problem
Relevance
The goal of research is to develop a technology-based solution to an important and relevant business problem.
o Estimate the Remaining Useful Life
o Recommendation for an investment o Asset reliability as least requirement Design
Assessment The utility, quality, and efficacy of a design artefact must be rigorously demonstrated trough well-executed evaluation methods.
o Clear output
o Creation of process overview Research
Rigour The research relies upon the application of rigorous methods in both the construction of the design artefact.
o A clear and well-funded process of coming to the design
solution Research
5. Design & Development
In this section the final solution of the design is presented. Before this final model is discussed, first the factors playing a role in the investment decision-making process are discussed. In chapter 6 the design criteria will be evaluated which will function as the design validation.
5.1 Main factors that play a role in the decision-making process
There are multiple factors that need to be taken into account in the investment decision-making process. These factors are based on literature and interviews with the stakeholders who were mentioned earlier in this paper. In this research, the investment decision-making process is defined as the steps that need to be taken to come to the best asset investment decision. To explain the factors that need to be taken into account in an asset investment decision-making process, the process is divided into two sub parts. The first part answers the question “When to invest?”, whereas the second part is concerned with the question “What to invest?”. The first part can be further divided into a data acquisition part and a modelling part, following the approach of Ghasemi and Hodkiewicz (2012). These authors made a model that presented the general steps to come to an asset’s RUL. In the second part, the best asset investment decision is based on a comparison between multiple investment options. These options represent different investment scenarios. The paper of Wang et al. (2015) discusses the three main investment scenarios. The three main scenarios identified by the authors are: continue operation, refurbishment and replacement of the asset.
First, the data acquisition phase will be discussed after which the modelling phase is further explained. Lastly, more details will be given about the decision making phase.
5.2 Data acquisition phase
The first factor included in this process is the gathering of data. To be able to establish the condition of an asset, data is required (Ghasemi and Hodkiewicz, 2012). Moreover, the data also is needed to determine the failure pattern of an asset (Schneider et al., 2006). The data can be gathered from two sources. The first source is the original equipment manufacturer (OEM). The second source is the data stored in the databases within the company. According to Si et al. (2011) there are two kinds of data that can be used in the RUL estimation: condition data and maintenance data.
5.2.1 OEM-data
31 maintenance activities will prolong its useful lifetime compared to the average. This is why it is important to transform the pre-set age (the time since the asset has been in operation) to the real age, which reflects the asset’s real condition (Kostic, 2003). 5.2.2 Condition data
To be able to analyse the condition of an asset, it needs to be monitored. The condition monitored (CM) data can be subdivided in direct and indirect CM data. Direct CM data describes the underlying state of the system directly like the wear and crack sizes. For indirect CM data is it only possible to indirectly or partially measure the underlying state of the system like the oil based monitoring of an asset. 5.2.3 Maintenance data
“Good maintenance data come from data that follow the business processes of sound maintenance solutions” (Campbell et al., 2011, p.105 ). Not only technical aspects need to be taken into consideration but also the historical data and its related costs to be able to make the best choices to ensure objectives are met in the long run (Louit et al., 2009 ). The collected maintenance data can be used as trigger for multiple occasions, among others: increasing maintenance costs, recurring maintenance activities and atypical maintenance activities. The main goal of the maintenance data is to indicate what happened and when (Campbell et al., 2011).
5.2.4 Useful data
5.3 Modelling phase
In this sub chapter more detail will be given to the modelling phase. This phase is composed of analysing the degradation pattern of an asset. Afterwards a boundary needs to be set to be able to estimate the RUL. First the technical RUL is established and afterwards also the economical RUL is analysed. This is done by analysing the future economical pattern and setting a threshold for the annual costs. 5.3.1 Degradation pattern The useful data, both the OEM data and the data collected at the company itself, can be used to establish an estimate on the degradation rate. The forecast can be plotted to get a degradation pattern. For rotating equipment, for instance, OEM data is often already sufficient to estimate the failure rate. These assets have a reliable degradation pattern if used according to OEM settings, e.g. a bathtub curve (Hashemian & Bean, 2011). However, for assets that have been tailor made, the company gathered data is more suited to predict the degradation pattern.
5.3.3 Boundary setting
Once the degradation pattern has been made, a boundary needs to be set in order to determine a certain condition level at which the asset should not be used and operated anymore. This boundary can be based on multiple factors. Most companies make use of risk matrices. In these risk matrices the chance of an incident is measured in combination with the impact on the company and its surroundings. The main five risk matrix dimensions are concerned with people, environment, assets (equipment/buildings), reputation and security. See appendix 1 for the general AkzoNobel risk matrices.
5.3.4 RUL estimation (technological)
Once the boundary has been set, along with the degradation rate, a RUL estimate can be made. The accuracy of the RUL estimation depends on how reliable and accurate data is as the estimate is based on this data. In the literature background chapter multiple models (physical, experimental, data-driven and hybrid) have been discussed that estimate the RUL, each having their advantages, disadvantages and special data requirements. Depending on the asset, data availability and environment, a RUL model is chosen to estimate the RUL. The estimated RUL needs to be compared with the desired RUL so the asset manager or investment decision maker knows to what extent the RUL needs to be prolonged. Although the technological RUL indicates at what point in time the asset is technically not viable, also an economical perspective on the RUL needs to be taken into account so establish when the asset is economically not viable anymore. 5.3.5 Future economical pattern
33 failures the corrective maintenance costs will increase. With the financial historical data, a forecast can be made of the total costs the asset will incur over its remaining lifetime.
5.3.6 Set a boundary (economical)
The total cost of ownership calculated for the current situation needs to be compared with a certain boundary to assess whether this figure is still within the limits. The limit is determined by a historical TCO prognosis that has been made when the asset was constructed or after the last large refurbishment. This historical TCO calculation can be used to decide whether the current situation is still within the allowable economic range. 5.3.7 Economical RUL After the economic forecast has been made and the boundary has been set, the economic RUL of the asset can be assessed. Most of the time, the technological RUL is longer than the economical RUL, which can lead to accepting the RUL on a technological perspective but disapproving the economical RUL compared to the desired RUL. If both RUL assessments yield a satisfactory result, no actions need to be taken. If, however, this is not the case, certain investment decisions need to be made. 5.4 Decision making phase
Assuming the estimated RUL is less than the desired RUL, action needs to be taken to make sure the asset can operate up till the desired RUL. Therefore, the last phase is the decision making phase. In the paper of Woodward (1997) life cycle costing was introduced. This method does not only take the initial purchasing costs into account but also considers other costs that a company incurs during the lifetime of an asset. The goal of this method is to maximise the value for money in the ownership of a physical asset. To minimise the total cost of ownership, the asset investment decision needs to be based on a thorough analysis to make sure the asset investment will result in the lowest overall cost on the long term (Ellram, 1995). The AkzoNobel asset managers and reliability engineers concur with including the lifecycle costs into the investment decision-making process. To be able to come to an asset investment decision, multiple investment scenarios need to be made. In the paper of Wang et al. (2015) the three main scenarios are continue production, refurbishment and replacement of the asset. In the designed model, the continue operation scenario has been called the “do nothing” scenario as this has a more clear meaning to it. After revising the investment scenarios on their total cost of ownership, the option with the lowest costs is chosen.
5.4.1 Do nothing scenario
5.4.2 Refurbishment scenario
The refurbishment scenario consists of a small investment where only a part of the asset is replaced or repaired. An asset is composed of multiple parts with each one having own specific failure modes and patterns. It might be economical more viable to repair or replace only specific parts of the asset instead of replacing the entire asset. The goal of a refurbishment is to increase the overall asset condition and/or decrease the OpEx in the following years. Figure 9 shows the result doing a refurbishment in respect to the asset condition. However, refurbishment does have some drawbacks. Firstly, most of the time the asset needs to be shut down to be able to perform the refurbishment, which can cause a loss production capacity. Secondly, the specific parts that need to be refurbished might be hard to reach which complicates the refurbishment (Schneider et al., 2005). On the other hand, a refurbishment only repairs or changes those parts that are causing the degraded condition and therefore is less costly than the replacement scenario.
5.4.3 Replacement scenario
In this scenario, the asset is proposed to be replaced. This means a new, like-for-like asset will be procured. Replacing an asset comes with certain advantages. Firstly, there is no issue with possible hard to reach parts to mend the asset as the entire asset is replaced. Secondly, a more efficient asset can be procured resulting in less operating and maintenance expenses. The main disadvantages on the other hand are the longer lead time, the high initial acquisition costs and the commitment to the asset.
5.5 Design solution
In this chapter the design solution is presented. The final design solution has been created according to all the design objectives while taking into account all the factors that need to be considered in an asset investment decision-making process, described above. The model can be seen in figure 15.
35 The “when to invest”-question is concerned with how pressing the current and future situation of the asset is and thus in what time window a certain investment decision needs to be made. The assessment and analysis of the asset’s technical and economical condition, its degradation pattern and economical forecast and the corresponding boundaries dictate the RUL estimation. The decision maker needs to assess the RUL estimate and compare it with the desired RUL. If the estimated RUL is less than the desired RUL (not satisfactory); mitigating actions need to be taken to prolong the RUL estimate. The three phases in the model are further explained below.
5.5.1 When to invest
First, the data is gathered from the different sources, OEM and internal databases. Next, the data is used to estimate the RUL on a technical and economical perspective. One of the main conclusions of the expert session was the fact that a distinction needs to be made between standard equipment and tailor made assets. This distinction boils down to the difference in available data. For the standard equipment like small pumps, the OEM has an extensive amount of data available. This gives the opportunity to use data-driven RUL models to predict the specific degradation pattern and the asset’s RUL. In figure 16 the most characteristic degradation patterns are shown. However, tailor made assets only rely on the data monitored at the company itself. Therefore, data-driven models are not suitable but a company needs to rely on its engineers. Consequently, for tailor made assets expert based models need to be used to estimate the degradation pattern. Figure 16 – Failure rate patterns (HPreliability) 5.5.2 What to invest
is used as a benchmark to assess the extent of advantages of the other scenarios. The comparison is made through the use of a TCO-analysis. The TCO-analysis can be computed in MS Excel where all the necessary costs can be put in for a time period between year 0 and 10. The items that are included into the TCO-analysis can be found in table 6. As an example in Appendix 2 an excerpt of an Excel file is added. Table 5 – TCO items per main category For each scenario, an estimated production amount can be filled out as well. This makes it possible to calculate the total price per unit product, which might aid in comparing the different scenarios, as they might not generate the same amount of product.
37
6. Evaluation
In this chapter the design criteria are reviewed to assess the solution design and evaluated on whether the solution complies with all the design criteria or not. The main criteria are concerned with the following items: design an artefact, problem relevance, design assessment, research rigour, and research contribution. Each of these criteria will be discussed below.
6.1 Design an artefact
The criterion to design an artefact has been met, as the solution to the managerial and academic problem is a decision-making model. This model is composed of three main phases forming a cycle and thus the output of the last cycle is used as the input of the new cycle. The model represents the decision-making process and encompasses all the necessary steps an asset practitioner needs to follow to come to a well-funded asset investment decision.
6.2 Problem relevance
The goal of design science is to construct an artefact aimed at changing the phenomena that occur (Hevner et al., 2004). The main phenomenon that the outcome of this research intends to change is concerned with the current asset investment decision-making process. The research has focussed on the different aspects of the decision-making process and thereby providing a solution to the problem. The stakeholder analysis resulted in three main problem related design criteria: estimation of asset RUL, recommendation for asset investment and asset reliability as a minimum requirement for the investment decision. 6.2.1 Estimation of asset RUL The modelling phase of the model is designed to assess the RUL of the asset. Next to the technological RUL assessment in this model also the economical RUL is incorporated. The incorporation of the economical RUL includes also the economical perspective in the final RUL assessment instead of only relying on the technological assessment. Most of the time the economical RUL is earlier than the technological RUL, which means continuing operation up till the technological RUL results in surpassing the optimal economical useful life span.
6.2.2 Recommendation for an asset investment
The right side of the model deals with the ‘what to invest’ question. The answer to this question is generated by assessing multiple investment scenarios. Different investment possibilities are reviewed and in order to get the best asset investment recommendation, a TCO-analysis is done.
6.2.3 Asset reliability
6.3 Design assessment
According to Hevner et al. (2004), the business environment has established the requirements upon which the assessment of the artefact is based. These assessment requirement are: functionality, completeness and usability.
6.3.1 Functionality
The goal of the model is to deliver a well-funded asset investment decision. By discussing a specific asset as an example, the steps of the model showed that the final output was clear and well-funded. The main arguments for the asset investment decision come from the thorough RUL analysis that estimates the time after which an asset requires a certain mitigating action. Also the economical RUL assessment is included in the model, whereas most methods look into only the technological RUL. This addition gives the model the required functions as it now incorporates a broader perspective of the asset RUL assessment. The design solution’s result is an asset investment decision that is yielded through a comparison between the investment scenarios. The TCO- analysis presents a clear support for the final lowest TCO investment decision.
6.3.2 Completeness
During the interviews and expert session, the model has been discussed. The design of the model was an iterative process, which after a couple of adjust-discuss cycles led to a complete, functional and useable model. The reliability engineer supported this model by stating the included steps depict exactly the way he would come to an asset investment decision. 6.3.3 Usability The usability can be linked to the extent the model provides an overview of the process, as a model with a clear overview of the different steps is easy to use. The separation of “when to invest” and “what to invest” gave a clear overview of the investment process. Moreover, all the steps have a clearly defined description of what the steps entail and what the outcome of the step will be to be able to move to the next step. By providing a clear overview and easy-to-follow steps the usability is guaranteed.
6.4. Research rigour
39 6.5 Research contribution
The design solution needs to provide a clear contribution in the area of the design artefact knowledge, which in this case is the asset management literature. The design needs to have some sort of novelty, significance and generalizability to be of contribution (Furguson, 2004).
6.5.1 Novelty
This decision model has integrated the asset RUL assessment with the asset investment scenarios. This way, the model describes the entire process from data gathering to best asset investment decision. To the best of the author’s knowledge, this type of integrated model has not been made and discussed in literature and thereby adhering to the novelty criterion.
6.5.2 Significance
The research is based on literature and on a case study. The main issue concerning significance is whether the model is based on enough data. The basis was formed with help from literature after which additions were made based on the AkzoNobel employees’ opinions on the topic. This was done through the case study where multiple employees have been interviewed on the aspects of the decision model. In addition, a discussion on the model was held during the expert session. Finally, the model was verified by examining an example in cooperation with the reliability engineers.
6.5.3 Generalizability
As the model has been verified at only one company, one could argue that the generalizability is limited. However, the main issue with the asset investment decision is that practitioners base their initial investment decisions on gut-feeling. This phenomenon has been researched and it was found that it is a common issue among capital-intensive companies (Bouyssou et al., 2006). Moreover, the model does not specify certain steps to the highest detail, which enables the model to be used for a variety of assets and even at other companies.
7. Discussion
The discussion of a research is composed of the following items: interpretation and explanation of the results, answering the research question, justifying the approach and limitations to the study. In the following sub chapters more detail will be given on each aspect.
7.1 Interpretation and explanation of results
The model designed in this study can be of great value to AkzoNobel and other companies. First of all, the model is designed to give an overview of the different steps that an asset practitioner needs to take to come eventually to an investment decision. The model is divided into a “when to invest”-part and a “what to invest” where the first part can be further divided into a data acquisition part and a modelling part. These sub divisions aid to the clear, logical and stepwise way the model is built up and result in a clear overview of the decision-making process. An asset should function to the characteristic it was designed for and in accordance to a yearly budget. In the modelling part, both asset aspects have been accounted for. Not only the technological RUL is assessed in the modelling part but also the economical RUL is taken into account. Therefore the asset practitioner can assess the RUL in a more broad and well-funded way. The reliability engineer supported this model by stating the included steps depict exactly the way he would come to an asset investment decision. Concluding, the entire model gives a well-funded argumentation to support the asset investment decision to convince the higher management of the criticality of the investment through the TCO analysis. 7.2 Answer research question This research set out to answer the following the question: How can a decision-making model be designed that supports the selection of the right asset investment decision, given a certain RUL assessment?
41 assessed by using the TCO. The TCO is composed of all the relevant cost aspects an asset incurs over its lifetime. The best scenario will yield the lowest TCO. 7.3 Justifying the research approach
The design science approach was used to conduct this study. The problem researched was of practical nature and therefore this methodology was chosen to produce a solution to a phenomenon that could be changed by implementing the produced artefact. The design solution was made partly by using the data and information gathered from the case study. This aided in the usability of the model as the AkzoNobel employees gave their input and professional opinion on the items and structure of the model.
7.4 Limitation of the study
In this sub chapter the limitations of this research are discussed. Also, recommendations for further research will be proposed.
8. Conclusion
The model presented in this study aids the asset management practitioner to make a well-funded asset investment decision that helps to support a request for (additional) funding to be able to make the necessary asset investment. The RUL assessment is composed of two parts: a technical RUL and an economical RUL. By assessing both these RULs of an asset, the best time to make an investment is established. The investment scenarios give the asset management practitioner the overview and insight it needs to make an important investment decision.
43
References
Abiteboul, S., Cluet, S., Milo, T., Mogilevsky, P., Siméon, J., & Zohar, S. (1999). Tools for data translation and integration. IEEE Data Eng. Bull., 22(1), 3–8.
Abu-Elanien, A. E., Salama, Abu-Elanien, A. E., Salama, M. M. A., & Bartnikas, R. (2011). A techno-economic method for replacing transformers. Power
Delivery, IEEE Transactions on, 26(2), 817-829.
Ahmadzadeh, F., & Lundberg, J. (2014). Remaining useful life estimation: review.
International Journal of System Assurance Engineering and Management, 5(4), 461–474.
Albee, S. (2005). America’s pathway to sustainable water and wastewater systems. Water Asset Management International, 1(1), 9–14.
Alchian, A. A. (1958). Economic replacement policy, Project RAND, RAND Corporation and Department of Economics UCLA, Santa Monica
Bagul, Y. G. (2009). Assessment of current health and remaining useful life of hard disk drives. Dissertation. Northeastern University, Boston, Mass.
Balsters, H., (2015) Lecture Slides of Course Design Methods, University of Groningen.
Banjevic, D., & Jardine, A. K. S. (2006). Calculation of reliability function and remaining useful life for a Markov failure time process. IMA Journal of
Management Mathematics, 17(2), 115-130.
Banjevic., D., (2009) Remaining useful life in theory and practice Metrika, 69, 337–349
Barlow, R. E., & Proschan, F. (1975). Statistical Theory of Reliability and Life Testing Probability Models To Begin With. Technometrics, 72, 304.
Bennett, P. (1996). ABM and the Procurement Cost Model. Management
Accounting, 77, 28–32.
Bethuyne, G. (1998). Optimal replacement under variable intensity of utilization and technological progress. The Engineering Economist, 43, 85–105.
Berg, H.-P., 2010. Risk management: procedures, methods and experiences. Reliab., Theory Appl. 1, 79–95.
