Decision Support Framework for Remanufacturing Process Planning: A Design Science Research

(1)

MASTER THESIS

Decision Support Framework for Remanufacturing Process Planning: A Design Science Research

MSc Dual Award Degree in

Operations and Supply Chain Management and

Technology and Operations Management

Author: Tom Joseph

Newcastle University: B6064634 University of Groningen: S3323463 Module Code: NBS8399/GRON116 Academic Supervisors

(2)

1

List of Figures

Figure 1 Remanufacturing process(Parkinson and Thompson 2003) ... 8

Figure 2 Phases in QFD Source: Crawford, 2008 ... 19

Figure 3Adaptation Logic ... 26

Figure 4 The regulative cycle (adapted from van Strien, 1997)... 27

Figure 5 The CBR Cycle ... 36

Figure 6 Parametric Search in SharePoint Source (Microsoft 2010) ... 42

Figure 7 Annotation of meta using term set hierarchy Adopted from Microsoft Office Support... 43

Figure 8 Attribute weigh description in DMS Source: Microsoft (2013) ... 45

Figure 9 Decision Support Framework for Remanufacturing Process Planning ... 52

(4)

3

Abstract

Remanufacturing process planning involves determining of reconditioning operations, its parameters and the order in which it must be implemented to restore the used product. Due to the uncertainty in the condition of returned products, process planning is ad-hoc, highly reliant on experts and require efficient decision-making process. Therefore, a decision support framework for remanufacturing process planning is hence proposed. Design Science Research methodology was utilized to design the framework. The framework was designed using Case-Based Reasoning techniques and Analytical Hierarchy Process to effectively reuse past remanufacturing experiences residing in the Document Management System. A novel attempt has been made to incorporate CBR techniques into DMS so that retrieval efficiency is enhanced. The framework designed was conceptually evaluated by conducting a case study. The proposed decision support framework helps to make sound and reliable process plans by efficient retrieval of knowledge which helps to reduce reliance on experts.

_____________________

Keywords: Remanufacturing, Process Planning, Case Based Reasoning, Analytical Hierarchy Process, Document Management System, Knowledge Reuse.

List of Abbreviations

AHP Analytical Hierarchy Process CBR Case-Based Reasoning

DMS Document Management System DSR Design Science Research

KMS Knowledge Management System NDT Non-Destructive Testing

OEM Original Equipment Manufacturer PSS Process Specification Sheet RI Repair Instruction

(5)

4

1 Introduction

Shrinking profit margins and increasing pricing pressure from competitors has led customers of capital-intensive equipment squeeze more revenue from their equipment investments. Replacement of the equipment which are nearing its end of life requires high capital investments. This has led customers insisting Original Equipment Manufacturers(OEM’s) to provide remanufacturing services which can extend the operational life of the used product considerably (Perella, 2014). Remanufacturing is defined as a recovery process by which products nearing end of life are completely disassembled following which its components are cleaned, repaired, reconditioned or replaced so that once assembled, these products performs its function as good as new (Mitra and Webster, 2008). In addition to customers’ demand, there has been mounting pressure from environmental legislation pushing OEM’s towards the path of remanufacturing since it offers a higher positive impact towards sustainability (Ferguson, 2010). Remanufacturing is often considered as the” ultimate form of recycling” (Xiang et al., 2011, p.680) due to its low carbon footprint(Fang et al., 2014). However, the motive behind providing remanufacturing services by OEM’s are widely contested. Profit and environmental legislation once considered to be main motives in providing remanufacturing services are replaced by increased customer orientation, warranty protection, branding and green image enhancement (Seitz, 2007)

(6)

5 Process planning is a decisive step which affects the success rate, cost and quality of the remanufacturing product. A systematic approach is therefore essential in the remanufacturing context so that quality uncertainties and dependence on experts is reduced. Jiang et al. (2016) states that the reliance on experts and the challenges due to quality uncertainties can be greatly reduced by applying knowledge management techniques. This is because process planners can quickly establish decisions on problems based on existing knowledge in knowledge management system(KMS) rather than relying on experts. KMS are heavily reliant on information technology and it “ facilitate the capture, storage, search, transfer and reuse of explicit knowledge in an organization”( Iskandar et al., 2017, p.70). In most firms, previous experiences and knowledge are conventionally stored as documents in the document management system(DMS) (Khan et al., 2015). These knowledges contained in the documents are meant to be effectively utilized for developing sound remanufacturing process plans. However, one of the major issues identified with DMS are efficient retrieval of relevant information(Damodaran and Olphert, 2000).This is a challenge for remanufacturing firms where past knowledge is stored in DMS as Jiang et al. (2016) states that efficient reuse of knowledge helps to create rapid and reliable process plans for remanufacturing. Moreover, effective utilization of knowledge rather than physical assets is the key to competitiveness for any firm as it can help the firms to gain competitive advantage (Moran, 1999).

1.1 Problem Description

(7)

6 towards the competitive advantage of the firm is subsumed in large sets of documents in the DMS. This makes it challenging for process planners to reuse experiences for making sound remanufacturing process plans as it is difficult to locate the relevant documents. In addition, since the firm deals with bespoke and one-off equipment, each documentation existing in the DMS are specific to each remanufacturing project conducted at the firm. That is knowledge regarding the current requirement might be implicitly represented in another project documentation. This requires process planner to troll through the documents in order to find the relevant documentation for the current requirement. It was also noted that there is a lot of tacit knowledge within the firm on locating the right documentation to base their decisions on. Retrieval of these documentations using the rudimentary search implemented in DMS results in an overwhelming number of documents, which may not be relevant to the current remanufacturing requirements. This leads to an overload of information among decision makers, which may cause difficulties identifying relevant information and therefore can lead to the risk of faulty decision-making.

In addition to above-mentioned problems, inherent uncertainties in the quality of used products makes remanufacturing process planning difficult and complex. Therefore, process planning at the firm is highly reliant on expert’s knowledge as they have to adapt their plans to varying conditions of the used product. The firm acknowledges the heavy reliance of “engineering expertise” as a major challenge and wish to reduce it. In conclusion, the firm wishes to 1) efficiently reuse knowledge to make sound process plans 2) decrease the uncertainties associated with process planning and above all 3) reduce the reliance on experts during process planning.

1.2 Research Aim, Objectives and Questions.

(8)

7 planning which aims to manage knowledge required for effective process planning decisions. Design Science Research (DSR) approach is used to theoretically design the proposed framework. This leads to the main research aim.

Design a decision support framework for remanufacturing process planning

The objectives of the decision support framework are multifold but are inherently interrelated to efficient knowledge reuse. The following are the identified objectives of the framework.

 Reduce the reliance on experts by enabling others to make effective decisions.  Efficient retrieval of knowledge from the DMS for making sound process plans and  Reduce the process planning uncertainty by suggesting proven solution.

To effectively design and evaluate the framework, an in-depth case study is required. Several supporting questions are therefore formulated to get deeper insights about the current remanufacturing process planning and how past knowledge are already reused in the firm.

 How is remanufacturing process planning carried out at an OEM?

 How are past knowledge reused for process planning decisions? And what

decisions are based upon it?

 What roles does IT systems such as DMS currently play to facilitate knowledge

reuse and what are the challenges associated with it?

Since the decision support framework have to take into account the uncertainties imposed by varying condition of used product, the following supporting questions are also formulated.

 What are the factors which contribute towards the uncertainties and challenges

(9)

8  What are the factors which affects remanufacturing process plans of components?

2 Literature Review

To fulfil the aim of designing a decision support framework, a literature review of remanufacturing organization characteristics, remanufacturing process planning and how knowledge is managed in project-based firm were carried out. Literature review helps to gain deep understanding about the current problem context and provides insights about prior design theories and frameworks used to solve similar issues.

2.1 Remanufacturing

Remanufacturing is an industrial process of recovering used products with desired specification and performance of” like-new “products (Wei et al., 2015). The process of remanufacturing involves numerous procedures to make the product perform as new as possible (Hatcher et al., 2011). Procedures include disassembly, cleaning, inspection, reconditioning/replacing components, re-assembly and quality testing of the final reassembled product(Wei et al., 2015). A schematic diagram of the processes involved in remanufacturing is illustrated in figure 1.

Figure 1 Remanufacturing process(Parkinson and Thompson 2003)

The operational process of remanufacturing starts with the complete disassembly of the used product in a non-destructive manner so as to ensure the reusability of the components (Fang et al., 2016). Disassembly is followed by thorough cleaning of each component which may include degreasing, dusting, de-rusting or even removal of surface coating such as paint(Parkinson and Thompson, 2003). Following disassembly and cleaning, thorough inspection of the components are carried out. Tang and Li (2012, p.53) states that these components “exhibits highly uncontrolled variability with respect to product condition, ranging from slightly used with minor blemishes to significantly damaged and requiring

(10)

9 extensive repair or replacement”. During the inspection phase, components are distinguished as reusable, re-manufacturable and non-remanufacturable part (Hammond, 1998). Reusable parts are components with good quality which can be reused in the remanufactured product with minimal or no reconditioning operations. Remanufacturable components have recoverable quality and therefore undergo thorough reconditioning process such as repairs, reworks etc. in order to be used again in the remanufactured product. Whereas non-remanufacturable components have poor quality and are often disposed-off and recycled. OEMs replace components which are disposed-off with new components. Replacement is the only alternative where repair is not possible and for large bespoke components, this is associated with “considerable replacement cost and prolonged schedule disruptions”(Bhaduri et al., 2003, p.397).Components which require reconditioning operations are further scrutinized using non-destructive testing methods, visual inspection and measurement of dimensions to ascertain the condition of the component(Parkinson and Thompson,2003).Defects and failures determined through the inspection process are rectified through reconditioning processes appropriate for each defect/failure. Reconditioning process include processes like machining, heat treatment or material deposition(Kin et al., 2014). Reconditioning is often considered as an essential part of remanufacturing as it restores used component to OEM’s original specifications (Parkinson and Thompson, 2003).In addition to bringing used products to their original specification, certain products are even re-engineered to accommodate design and technological enhancement so that the remanufactured product has better performance than the original product(Du et al., 2013). Levine (1993) points out that product enhancements are mainly based on the already proven and existing knowledge residing within the firm. According to Fang et al. (2014), disassembly and component reconditioning are the most important process which determine the success of remanufacturing.

(11)

10 performance with a warranty that is equivalent or better than that of the newly manufactured product” whereas refurbishment aims to “return a used product to a satisfactory working condition by rebuilding or repairing major components that are close to failure” (BS8887-2-2009, (BS8887-2-2009, cited in Bakker et al., 2014, p.11). Remanufacturing offers several advantages when compared to refurbishment. It provides superior products and maximum resource utilization which reduces environmental impact and energy consumption. Parkinson and Thompson (2003,p.255) states that remanufactured product, when compared to refurbished product, has the potential to be “updated and enhanced” utilizing cutting-edge technology available in the domain.

Generally, not all products are suitable for remanufacturing. Lund (1996) identified the typical characteristics of products that are ideal candidates for remanufacturing. These products possess a non-consumable core which are tough and durable and have slow product obsolescence. Low et al. (1996) identified that products having components of high value and low cost for recovering it to the original state are suited since the primary raw material used for remanufacturing are used components. These include high valued products such as turbines, aero-engines which have high valued components. However, Guide et al. (2003) points out that even though remanufacturing industry is concentrated in high valued products which typically are low volume operations, the industry is also shifting towards the remanufacture of low-value products which has higher obsolescence rate. Guide and Wassenhove (2001) attributes the shift due to environmental legislation imposed upon the manufacturer so as to make them responsible for preventing discarded products ending up in landfills.

2.2 Characteristics of Remanufacturing Organizations

(12)

11 must be taken into account. ETO products have components designed specifically to the customers requirement which could include other standard components (Bozarth and Handfield, 2006). Guide (2000) points out that ETO products usually require the firm to undertake remanufacture to order1_{(RMTO) product position strategy because the customer} requires the same unit returned because of its customized components. The paper also pointed that unlike firms adopting remanufactured to stock2_{(RMTS) strategy, the uncertainty} and complexities associated with balancing used product returns and remanufactured product demand is not present in firms having RMTO strategy. This is because customers act as the provider of used product and the purchaser of the same remanufactured product. Lund (1996) identifies RMTS strategy as batch processing operations whereas RMTO strategy as one at a time operation. The paper points out that RMTO strategy is economically viable for high valued products as it can absorb the associated increase in labour cost than low valued products. Guide et al. (2003b) conducted an elaborate study regarding different remanufacturing strategies (RMTO, RMTS and reassemble to order(RATO) remanufacturing strategies) and their associated managerial concerns. The study found that RMTO firms deals with products having high complexity which make remanufacturing planning complex due to the “variable quality of returns, large number of parts, and variable remanufacturing requirements for each part” (Guide et al., 2003b, p.272). Furthermore, RMTO firms when compared to other RMTS and RATO firms has 1) higher variability in return quality due their high product complexity,2) low uncertainty in the availability of raw material (used product) due to low volume customer-centric operations and 3) higher remanufacturing complexity which requires product components to go through varied and extensive reconditioning operations due to products complexity. Guide (2000, p.476) states the dependence on customer-owned asset makes RMTO strategy of having “no risky core acquisitions based on projected demands”. Therefore, RMTO firms does not require the extensive management the inventory of used and remanufactured products as in RMTS and RATO firms (Ostlin et al.,2008). Most RMTO firms have service contract relationships and therefore product returns are forecastable with some degree of certainty. This allows production schedules to be made

1_{RMTO is similar to make to order strategy where remanufacturing is carried out on customer owned asset}

and upon completion are returned to the customer.

2_{RMTS is similar to make to stock strategy where used products are bought from the customer based on}

(13)

12 far in advance (Ostlin et al.,2008; Guide et al., 2003b). Therefore, RMTO firms have many operational advantages compared to RMTS firm as the uncertainty regarding the quantity, timing of used products and the demand rate is almost irrelevant due to low volume customer-centric operations. Even though RMTO firms have little uncertainty in return and demand rate, Guide and Wassenhove (2001, p.144) points out that the “primary reason for complexity is the high degree of variability in the quality of used products that serve as raw materials for the production process”. Other challenges associated with RMTO firms are long procurement cycle of low volume and customized parts (Guide, 2000) and the customers requiring the firm to return the remanufactured product within the shortest time frame (Ostlin et al.,2008). These challenges have direct implication on the management of replacement part for the used product (Ijomah,1999). For combating long procurement cycles and to have desirable service levels, Guide and Srivastava (1997) suggest having optimal amount of inventory balancing the high cost of replacement parts. Since the primary concern of RMTO firm is to restore the variable quality products, the following section will address the operational challenges associated with process planning.

2.3 Complexity in Remanufacturing Process Planning

(14)

13 and flexible to suit the component and therefore making it highly reliant on experts knowledge(Du et al., 2013). Added to that, the sequence of operations strongly depends upon the condition of the used product thus resulting in unfixed process routings(Tang and Li,2012). Therefore while process planning, uncertainties caused due to varying quality must be incorporated and used as an important input in process planning decisions (Aksoy and Gupta, 2011). Zhou et al.(2014) points out that changes in some factors causes corresponding changes in process parameters eventually leading to different process plans. These factors that determines the remanufacturing process plans are collectively termed as influence factors. The different influence factors as identified by researchers(Jiang et al., 2016; Zhou et al., 2014; Kin et al., 2014) are stated below in Table 1. Each of these influence factors listed in Table 1 may affect process plans in varying degrees.

Table 1Factors affecting remanufacturing plans

Influence Factors

Failure Characteristics Failure root cause, Failure Type, Failure location, Failure degree

Defect Characteristics Defect root cause, Defect Type, Defect location, Defect degree

Component Characteristics Component Type, Geometrical characteristics of component, Machining allowance (Precision required, Tolerances, Error permissible, Roughness factor).

Material Characteristics Material Type, Material Properties (Hardness, Corrosion Resistance, Wear Resistance, Heat Resistance)

(15)

14 sustainability (Ferguson and Souza, 2016) However, Jiang et al. (2016,p.3) states “there is little research aiming at the process planning for restoration/reconditioning” of used products.

Kin et al. (2014) analyzed the conditions of used product to determine and plan the process operations of components. The reconditioning process which is a subset of remanufacturing operations were determined through analyzing the components quality condition. The reconditioning operations were ordered through ranking the criticality of defects inputs using failure mode and effect analysis. Defects with higher rank priority was given preference in reconditioning process which ensures quality in the remanufactured product. However, the methodology proposed was a conceptual in nature and does not provide how optimization of the selection process is carried out. Similarly, Parkinson and Thompson(2004) presented a systematic approach to remanufacturing process planning based upon the failure mode and effect analysis(FMEA) method.The study concluded that using FMEA outputs, the remanufactured products can be made as reliable as new products. Guide et al. (1997) analyzed quality uncertainties in used product and determined how it affected the remanufacturing inventory and production plans. Similarly, Tang and Li,(2012) studied the impact of uncertainty in remanufacturing and how it can be managed.The study investigated the uncertain quality of returns and how it affects the production planning and schedules. Fang et al. (2016) developed a holistic framework for assessing the feasibility of a product to be remanufactured. Feasibility was determined through building a knowledge base from 3D Computer Aided Design (CAD) models. The framework also entails a systematic approach from to generate an optimal disassembly route for products which was derived using part accessibility metric. This metric contains information about degrees of freedom of each component in the product and disassembly was based on decreasing order of degrees of freedom. Process planning sequence was determined through failure modes and the design information that was extracted through exploded 3D CAD models.

2.4 Knowledge Transfer in RMTO Projects

(16)

15 are defined in the literature in diverse ways, however a consensus exists that projects are a temporary, time constraint set of activities having a defined and distinct objective. According to Slack et al. (2010), the characteristics projects have in common are - an objective typically in terms of cost, quality and time, have defined start and end dates, are unique and temporary with some degree of complexity and uncertainty. Due to the specific characteristics of the projects; knowledge transfer across the firm is challenging (Landaeta, 2008). According to Hallgreen and Maaninen-Olsson (2005); a lack of key knowledge in a project; risks accomplishing the objectives due to incompetence to deal with issues and problems. Often projects fail due to the lack of lessons learned or lack of knowledge transfer from the project teams(Duffield, and Whitty,2016).Whereas effective knowledge transfer would increase the project capability, performance and reduces the uncertainty and ambiguity of projects. Barquet et al. (2013) emphasized the requirement of remanufacturing to have employees that are qualified to deal with the type of product and its complexity. Similarly, Hermansson and Sundin (2005) stressed the importance of firms to uphold a broad competence level among its employees so as to remain flexible to deal with uncertainties related to each stage of remanufacturing operations. These reiterates the importance of effective organization-wide knowledge transfer among the employees involved in remanufacturing projects. Effective knowledge transfer among projects helps firms to build upon competencies and skills attained during the conduct of projects and prevents repeating action and decisions that caused problems (Disterer, 2002). It also prevents projects from ‘reinventing the wheel’ (finding already known solutions) thereby reducing the cost of duplicating efforts to reinvent it (Schacht and Mädche, 2013).

(17)

16 enabled through codification of knowledge in documents and personalization mechanism. Personalization utilizes person to person knowledge transfer mechanism in which individuals having the flexibility to re-negotiate meaning and adapt knowledge according to the context (Hansen et.al.,1999). It is often seen as ad-hoc and informal (Boh, 2007). Even though personalization allows transference of contextualized knowledge, it often seen as management of communication rather than a knowledge management technique because knowledge is not collected, edited or documented (Disterer, 2002). Knowledge codification, on the other hand, involves codification of explicit knowledge in central databases (lesson learned reports, project profile reports, design templates, manuals etc.) and is highly reliant of efficient information technology to transfer and share knowledge across project (Scarbrough et al.,2004; Hansen et.al.,1999). Codification of knowledge helps to achieve scale in knowledge reuse as large amount of knowledge can be stored and large number of employees can search for and reuse knowledge (Hansen et.al.,1999). Compared to personalization strategy, it allows project to store and transfer of knowledge across space and time which allows future projects to transform the underlying knowledge (Foray and Steinmueller; 2001).

(18)

(19)

18

2.5 Relevance of Customer Input in Remanufacturing

Parkinson and Thompson (2004) states that remanufacturers need to mandatorily consider customer requirements, specifications, legislation and other standards for remanufactured product. A customer-facing approach is a key competitive strategy which allows businesses to provide its customers with “products or services that are more in tune to their requirements and their view of quality” (Shahin and Nikneshan, 2008, p.70). Moreover, service-based remanufacturers must maintain close relationships with their customer throughout the delivery of services in order to reduce the risk and the perceived relationship costs (Ostlin et.al.,2008). Delivering value is a cornerstone of any relationship and is not just based on the cost benefit offered but has various dimensions which involves the quality of products and services, reliability, speed, ease of use, responsiveness, service excellence etc. (Shahin and Nikneshan, 2008). One of the most important aspect of customer relationship management is to understand the customer requirements and match it with products and services provided by the firm (Nawaser et al., 2014).

(20)

19

Figure 2 Phases in QFD Source: Crawford, 2008

(21)

(22)

21

3 Methodology

This section will describe the research method and the “conceptual structure in which the research is conducted” (Kothari,2004, p.31).

3.1 Research Design

The firm in the study faces several challenges in remanufacturing process planning and has difficulty in reusing the knowledge residing in DMS while making planning decisions. The research aims to alleviate the aforementioned challenges by constructing an IT-oriented decision support framework which can aid process planning decisions by effectively reusing knowledge from the incumbent DMS. Design Science Research (DSR) approach is used to design the proposed framework. The rationale behind using this approach is that; DSR is a pragmatic research paradigm which involves the designing of an innovative artefact to solve a real business problem (van Aken et al.,2016). In addition, DSR is highly relevant to information systems research and the research community has recognized the “importance of DSR to improve the effectiveness and utility of IT artefact in the context of solving real-world problems” (Hevner and Chatterjee, 2010, p.9). An artefact can take the form of a construct, model, framework or an instantiation for a specified problem domain (Hevner et al., 2004) where a framework is described as a “real or conceptual guides to serve as support or guide”(Vaishnavi and Kuechler 2004, p.12). Eventhough Hevner et al.(2004) and March et al. (2004) emphasises about technology-based artefacts represented by software or formal logic, they are also open to an exploration of organisations, policies, and work practices as designed artefacts (Baekgaard, 2015).

(23)

22 Chatterjee (2010, p.15) classifies ‘design as research’ as a distinctive class of DSR aligned to organization and information sciences where the research project is performed in a specific application context within which the “resulting designs are influenced by the opportunities and constraints of the application domain”. Whereas ‘researching design’ largely focuses on methods of designing in the fields of architecture, engineering and product design and are largely domain independent. Since ‘design as a research’ paradigm has a high priority in the application domain; the structures of the application domain are inscribed into artefact.Sein et al. (2011,p.38) states that design efforts and contextual factors of the application domain becomes manifested in the “form, structure and conceptualization of the artefact”.The conceptual artefact thus generated require ongoing refinement in the application domain through evaluation methods.

DSR research has been previously utilized in operations management field problems. For example, Trovinger and Bohn (2005) utilized DSR research in reducing the capacity loss in a PCB assembly line. Capacity loss was caused due to large setup time. The paper aimed to reduce the setup time using Single Minute Exchange of Dies (SMED). Here, SMED which was predominantly developed for manufacturing metal components was contextualized to be used in PCB assembly. In addition, various information technology tools were utilized for handling the vast complexity in set up process. Baloh and Desouza (2009) used design science research to design a conceptual knowledge management system(KMS) model. With the model created, the researcher aims to solve the problem context - “how can they stop employees from reinventing the same or even suboptimal solutions to problems that were already solved by someone else” (p.2). The KMS model was designed upfront in a positivist manner and was conceptually evaluated through an exploratory qualitative case study which involved discussion of the model with the practitioners. Similarly, Schacht and Mädche (2013) developed a KMS for increasing the reuse of project knowledge among various projects using DSR. The above-mentioned problem contexts strongly correspond to our problem context and therefore the use of DSR can be justified.

3.2 Research Characteristics of Design Science Research

(24)

23 problems”(Vaishnavi and Kuechler 2004, p.12) whereas, DSR addresses to solve unique problems in more “effective or effective ways” (Hevner et al., 2004, p.81). In this research, the problem context is given, and a decision support framework is designed to improve the problem context. DSR therefore takes a pragmatic stance with the final research product being a design which have pragmatic validity(van Aken et al., 2016). Pragmatism has a practical and utilitarian philosophy (Teddlie and Tashakkori, 2003) and the artefacts are “assessed against criteria of value or utility”(Järvinen, 2007, p.51).This is supported by Hevner et al. (2004,p.12) who states that design science research must be “evaluated in light of its practical implication” along with the rigour with which the artefact is created. Rigour involves grounding the designing process on already existing design products and processes considering scientific theories and methods. It also include “providing clear rationales for the selections of design methods ”(Gregor and Hevner, 2013, p.350).

Hevner et al. (2004,p.98) differentiates behavioural science paradigm and design science paradigm. Behavioural science “seeks to find what is true” whereas design science paradigm “seeks to create what is effective”. van Aken et al. (2016,p.6) states that even though a research takes a DSR approach, it has a “explanatory and design component”. The design part creates improvement-oriented solution and the explanatory part provides a comprehensive foundation of the field problem. In this research, the design component denotes the construction of the decision support framework which is based on the initial problem representation. Whereas the explanatory component explores opportunities and constraints in the application domain where in which the remanufacturing challenges and knowledge reuse issues are further explored. Here the awareness of the problem is revisited. The explanatory face also involves an evaluation phase in which the designed framework is conceptually “assessed against criteria of value or utility” (March and Smith 1995, p.44). This approach is a part of iterative development cycle where further explanatory knowledge regarding the application domain and the framework serves a formative part of the design phase in the next iteration of the design cycle (Gregor and Hevner,2013).

(25)

24 (2004,p.23) noted that “design science research is very similar to the action research methodology of the interpretive paradigm” as observations made by the researcher are interpreted throughout the interventionist cycle. This research falls largely in interpretivist paradigm as it mainly involves qualitative analysis of the application domain which would serve as input for further DSR cycles.

3.3 Research Strategy

DSR is a “domain independent research strategy” which essentially consist of the descriptive/explanatory and design part(van Aken et al., 2016, p.8). For conducting DSR initial cycle, design science research methodology(DSRM) prescribed by Peffers et al. (2007) was utilized. DSRM serves as a guideline for this research and it essentially consist of four core activities. DSRM was used throughout this research and the research paper is structured accordingly.

Activity 1- Problem identification and motivation: The first activity of DSRM consist of

identification of the problem faced by the organization and having a general awareness of the same. Pilot interviews were conducted at the organization to create an awareness of the problem. The problem identified will help guide towards the proposed solution(Peffers et al., 2007). In this research, the section 1.1 describes the problems and challenges faced by the organization and motivates the researcher to pursue a solution to improve upon the context. A critical review of the literature was also conducted to further understand the context of the problem.

Activity 2- Define the objectives for a solution: The objectives for the solution are

inferred rationally from the problem identified. Objectives state how the framework produced from DSR is expected to improve upon the problem context. In this research, section 1.2 describes the how the decision support framework is expected to support the solutions to the current problem context.

Activity 3-Design and Development: Design and construction of a new artefact is the

(26)

25 support framework) and involves sub-activities such as “determining the artefact’s desired functionality and its architecture and then creating the actual artefact”(Peffers et al., 2007, p.13). Since the research carried out is in the first phase of DSR cycle, the problem and objectives identified in activity 1 and 2 only “provide a platform for generating the initial design of the IT artifact, which is further shaped by organizational use and subsequent design cycles” (Sein et al. 2011,p.38).Therefore a conceptual framework is proposed in this research. Baekgaard (2015) argues that conceptual model of artefacts have its relevance in DSR and illustrates its potential to solve a problem context.Since the solution is a design, there are many artefacts which can be designed to solve the problem context. Out of designed artefacts, an optimal solution can be chosen to improve/solve the problem context. However, this strategy is not feasible in our current DSR due to “the sheer size and complexity of the solution space”(Hevner et al., 2004, p.89).Therefore the principle of satisficing was applied to obtain satisfactory solutions(Simon, 1996). It involves applying heuristics search strategies to produce an tentative artefact which provides satisfactory solution to improve current problem domain (Barr et al., 1995, Hevner et al., 2004).Tentative artefacts are suggestions for a problem solution and are“ abductively drawn from the existing knowledge/theory base for the problem area”(Vaishnavi and Kuechler, 2004, p.10).These suggestions have knowledge gaps and are therefore further developed to design an artefact using the rigour cycle, i.e. utilizing “prior theory and existing design knowledge”(Hevner et al., 2004, p.89). Rigor cycle ensures innovation in the designed artefact by drawing from a knowledge base of scientific theories, methods, existing processes and artefacts (Hevner and Chatterjee, 2010).

(27)

26

Figure 3Adaptation Logic

The existing body of knowledge “provides the raw materials from and through which DSR research is accomplished”(Hevner et al., 2004, p.80). In this research, the artefacts which was used from the existing body of knowledge are operation management technique such as Analytical Hierarchy Process(AHP) and artificial intelligence technique such as Case-Based Reasoning(CBR) to help aid process planners in decision making. Designing the framework involved improving upon the current DMS by incorporating CBR techniques in conjunction with analytical hierarchy process(AHP) all by thoroughly using references from the knowledge base (Rigor Cycle).

Activity 4- Evaluation and demonstration: It involves demonstrating how the artefact

(28)

27 reflect about the fit/misfit between a set of requirements and one or more designed solutions”. van Strien’s (1997) regulative cycle is utilized to conceptually evaluate the framework (see figure 4) Pragmatic validity was evaluated by conceptually analyzing whether the designed artefact works and the desired outcome is obtained after contextualization(van Aken et al., 2016).

Figure 4 The regulative cycle (adapted from van Strien, 1997)

(29)

28 be further developed and implemented in practice. At the end of evaluation phase researcher can decide whether to iterate back to design and development or can leave the improvement needed to subsequent research projects(Peffers et al.,2007).

Activity 5-Communication: The final artefact created as a part of DSR must be

effectively communicated to researchers who can extend and study them in the problem context and to practitioners who decides if it is to implemented within the firm(Hevner et al., 2004).

3.4 Setting of the Research

The research is conducted at an OEM based in the UK with a distinctive service orientation. The focal point of the research is the engineering department within the firm where remanufacturing process planning is carried out. A single case was studied was conducted at the firm. Siggelkow (2007) points out that a single case can richly describe the existence of a problem.

3.5 Data Collection

According to van Aken et al. (2016), DSR does not require specific data gathering and analysis methods when compared to other research strategies. Data collection was carried out via two qualitative data collection method, multiple interviews and content analysis. Multiple sources of evidence ensured the construct validity of the research (Karlsson, 2009). McCutcheon and Meredih (1993) also noted that it improves the validity and reliability of the research. The following section will elaborate on the procedures undertaken while collecting data.

(30)

29 the unstructured interviews, pure and confrontive inquiry techniques was used. Pure inquiry techniques involve eliciting from the interviewees about “what is taking place and listens carefully and neutrally” (Karlsson, 2009, p.256). Confrontive inquiry technique involves researcher sharing his ideas regarding the solution [decision support framework] to the current problem/challenges thereby eliciting from the interviewees what they think about the proposed solution (Karlsson, 2009, p.256). Multiple respondents who are associated with remanufacturing were interviewed. Respondents consisted of a diverse set of people having different job roles and function. This included interviews with the engineering team, business, quality and the non-destructive testing (NDT) team. The interviews with business team were primarily to understand the initial problem context. The interviewees list is attached in Appendix 8.2. As multiple perspectives of interviewees increase the construct validity of this study. Prior to the start of the interview, participants were informed about the context being researched and consent was taken to record the interview. In addition, an interview protocol was followed to increase the reliability of the research. Face to face interviews were conducted in a quiet environment. The recorded interviews were manually transcribed in order to ensure higher quality transcripts. Recording and transcribing the interviews also helped to prevent observers bias and thus ensures internal validity (Karlsson, 2009).

Content Analysis: Taking into consideration company’s data privacy and protection policy, the content analysis of documents was carried out. The firm store its historical experiences in remanufacturing as repair instruction and lessons learned in the firms’ DMS. The contents of these documents were studied in order to gain critical information about how past knowledge is captured in the firm and what it entails. Analysis of the structure in which these documents were stored in the DMS gave insights about why knowledge reuse was challenging for process planners. Content analysis also included firm’s information leaflets, Internal white papers and marketing brochures.

(31)

30

3.6 Data Analysis

Interviews yield unstructured text-based data which needs to be coded for facilitate data analysis. In order to aid data analysis, interviews were first transcribed manually (See Appendix 8.5). The transcribed interviews were then coded by subdividing and categorizing data using tags to narrow down and filter large amount of data (Karlsson, 2009). Coding also served as a platform to integrate all data collected from multiple data collection methods such as content analysis and informal discussions (Saunders et.al., 2012). At first, open coding was utilized to narrow down on distinct concepts and categories found in the data. These were coded using priori codes (start list of codes) and emergent codes. Priori codes was based on the investigated problem domain. After the initial coding, axial coding was carried out. The purpose is to strategically reassemble data that were “split” or “fractured” during the initial coding process (Strauss and Corbin, 1998, p. 124 cited in Saldaña 2015). It involved sorting and grouping the initial codes assigned which were then regrouped into sub-categories and subsequently categorized into main topics. The codes were also relabeled to fit them into conceptual categories (see Appendix 8.3 and 8.4). The coding process was cyclic in nature as it involves linking data to categories and back to other data(Saldaña,2015).

3.7 Limitations

Multiple data collection methods which involves a mix interviews and questionnaires could have been incorporated adding to construct validity of our research (data triangulation). The artefact was evaluated conceptually without the implementing it in the problem context. Therefore, the regulative cycle of van Stiren (1997) (shown in figure 4) was utilized to due to limited time frame of the study. Conceptual evaluation could lack empirical evidence on whether the artefact would fulfil the required objectives in the problem context. Furthermore, only one iteration of design science research was carried out. Further iterations in designs are required after the evaluation of the artefact in the problem context. In addition, the artefact was designed and developed on the problem context of single firm, thereby limiting the generalizability of the DSR artefact.

3.8 Ethics

(32)

31 Ethical and moral aspects was considered throughout the conduct of the research. The researcher respected and acknowledged the intellectual property rights of the case company when content analysis was done. A confidentiality agreement was signed with the organization prior to the start of the research. Consent to use the data for further analysis was taken from each interviewee before the start of the interview. Participants were also ensured confidentiality and privacy during the entire research process and in research reports. Interviewees rights were respected throughout the interview by allowing them to refuse to answer any questions or to withdraw from the interview at any point of time. Each participant in interview was also informed about the aim of the study beforehand the interview. To ensure confidentiality and anonymity of participants in the research report, all the participant was addressed using pseudonyms in the findings and discussions and evaluation section. As a research student at the university, the universities code of ethics was also followed, and the ethics review was conducted during the research proposal phase.

4 Conceptual Design of Decision Support Framework

As per design science guidelines, the framework must be constructed strictly based on existing research and the assessing of usability and utility is done in subsequent evaluation phase (van Aken et al.,2016). The literature review conducted in the previous sections gave more insights about the context of the problem. The following section aims to construct the decision support framework(artefact) which is relevant to the problem context explained in the section 1.1.

(33)

32 judgement; process planner requires to acquire and assimilate vast amount of information due to product variability and quality uncertainty which puts a” premium on process planners to make efficient and effective decision making” (Ijomah, 1999, p.163). The firm in this research have documentation which contain project experiences regarding remanufacturing process planning; however, process planners find it difficult to find the appropriate information among the large sets of documentation. This would hamper the decision-making process as the quality of decision making decreases beyond a certain amount of information. This is because heavy information load will confuse the decision maker’s ability to make accurate judgements (Chen and Harrell cited in Eppler et.al, 2004). This can be avoided by using artificial intelligence(AI) systems or knowledge-based methods that reduces a large set of options to manageable size. Application of AI techniques assures that the information available to the decision makers are of high value, delivered in the most convenient format (Eppler et.al, 2004).

(34)

33 techniques are therefore adopted to design the decision support framework as it embodies efficient knowledge retrieval and because the firm already has a library of past remanufacturing project experience which can act as solution to future decision problems.

Moreover, CBR is widely used as a decision support tool where decision makers can identify previous situations that matches the current situation and most likely the decisions made for the previous situation can act as solutions to the current decision problem( eg. Chow et al.,2006). Therefore, it allows decision makers to provide solutions to problems quickly and propose a solution in an area which is not completely understood by the decision maker (Kolodner, 1993).CBR also qualifies as a knowledge management (KM) methodology since it support all the KM activities such as knowledge capture, retention and reuse and effectively contributes to the knowledge sharing culture. However contrary to KMS; CBR techniques are only effective in accomplishing a specific task in a specific domain rather than variety of tasks in different domains (Watson, 2001). The following section will describe about CBR and its underlying concepts. Discussion on how CBR will be utilized in the current problem context will be explained in further sections.

4.1 Case-Based Reasoning

(35)

34 formality, ontologies can range from a simple taxonomic hierarchy of classes to higher-order logic (Bergmann and Schaaf 2003). It aids in reuse of knowledge by exploiting the formal specifications of terms and relationships specified in a given domain (Gruber, 1993, cited in Noy, N.F. and McGuinness, 2001).

(36)

35 conditions of the used product. To solve this issue, Jiang et al. (2016) devised a set of rules to determine the corresponding influence factors. Retrieval of relevant cases were based on CBR techniques by computing the similarity between influence factors. Rough set was utilized to increase the accuracy of case retrieval by removing redundant influence factors which does not aid in accurate retrievals. With respect to transferring project experience, CBR has proved its worth to managing lesson learned across a project-based company (Friedrich et al., 2002), to generate project plans from past plans (Xu and Muñoz-Avila,2004) and for sharing project experiences(Dorn,2016). However, the study conducted by Dorn (2016) assumes that codified project experiences can be efficiently transferred across project by retrieval based on project attributes such as project type, customer type or project name. Weber et al. (2001) points out the relevance of CBR techniques in organizing lessons learned documents in project-based firms. These documents can be represented in the system for processes where lessons are applicable. This allows retrieval based on applicability and in the context where they are delivered. Based on this, Weber and Aha (2003) created a CBR system for lessons learned to promote retrieval based on applicability and demonstrated its advantages in a military operations lesson learned system. The system also has proactive distribution methods which allows the system to disseminated knowledge when required by monitoring the users’ context. However, proactive distribution requires the integration of automated tools such as an ERP system for monitoring its users. Before moving into how CBR can be utilized in the problem context, the main concept and approaches in CBR are discussed in following section.

4.1.1 CBR Cycle

(37)

36

Figure 5 The CBR Cycle

Case Retrieve:

The retrieval phase plays a pivotal role in the problem-solving cycle in CBR since it retrieves similar cases to the target case3_{which effectively should have similar solutions. Retrieval of} the case involves fetching similar cases solved in the past by parsing and checking the similarity of the current problem query with the problem part of existing cases residing in the case base4_{(Lau et al., 2009). Problem queries can be inputted through free text box query or} using parametric search boxes in CBR systems. In free text box entry, the users are free to formulate their query, but search results will be bound to have ambiguity as free text searches won’t be able to capture every aspect of user’s problem. To avoid ambiguity, Ruiz et al. (2013) suggested a step prior to the retrieval phase to elaborate the problem query with a semantically derived list of core descriptors that is derived from the domain ontology. It involves filling in and filtering the raw description of the problem query based on domain ontology.An alternative to this is parametric search boxes, where user can input their queries in multiple search boxes containing structured drop-down boxes, radio buttons, switches etc. Here, the problem query can be captured to a detailed level to describe the target case. However, Landau (2010) argues that parametrized search might confuse and overwhelm

(38)

37 inexperienced end user as it is more oriented towards expert end users and suggests the use of a mixture of free text interaction5_{and parameterized search for inputting problem queries.}

The basic approach involved in computing the similarity between the target case and the source cases6_{is by computing the global similarities. Here, similarity between the cases} are computed pairwise for each attribute in the “problem description part” to obtain the local similarity (Haque et al., 2000; Liao et al., 1998). It should be noted that not all attributes in the target case are equally important to the current problem context (Aamodt and Plaza,1994). Therefore, weights for each attribute in the “problem description part” of target case should be assigned by the user. Assignment of attribute weights helps CBR systems in identifying the attributes which are more important in the current problem context. This results in increased retrieval accuracy as CBR systems can check the similarity between cases by taking into account the weight of each attribute (Tamoor et al., 2015). Local similarity hence obtained are then combined to form global similarity. Source cases are then ranked according to the global similarity computed. From the retrieved cases, the highest ranked source case solution may be reused to solve the current problem.

Case Reuse:

After relevant case has been retrieved; the next step in CBR is case reuse. The reuse phase consists of copying or adapting the solution from the retrieved case to match the target problem. There are two possibilities in case reuse. If the target case and the source case’s “problem description part” have high similarity and the differences are irrelevant, the whole or the part of the solution of the source case can be reused as a solution to the target case. If else, the solution of the target case must be adapted to consider those differences (Kolodner, 1993). Adaptations are of two types namely derivational reuse and transformational reuse. Derivational reuse involves reusing the method in which the past solution was created whereas transformational reuse involve adapting the solution of the retrieved case for reuse (Aamodt and Plaza, 1994). Transformational reuse is a knowledge-intensive task as the retrieved case is not a direct solution to the target case. Therefore, it involves adapting the solution from the retrieved source case to solve the target problem. Transformational

5_{Free text interaction allows user to query the case base using free text thereby allowing end users to}

formulate problem in their own word.

(39)

38 Adaptation can be done manually and automatically (Weber et al., 2005). However, except in highly limited domain, there is little evidence to support automatic adaptation methods in complex domains are reliable.Therefore, in complex domains, adaptation is done manually and heuristically with the help of experts in the domain(Aamodt and Plaza, 1994).

Case Revise and Retain: Case Revise and Retain is the last step in the CBR cycle where new

cases are added to the case base which leads to continual improvement of knowledge base of the firm. Revise step involves evaluating and verifying the solution after implementation and is presented as confirmed solution to the new problem case (Kolodner,1993). After the revise step, the new cases are stored in the case base after some processing and filtering to comply with the CBR structure. This involves structuring the case as problem and solution part. As more cases are added, there is better coverage of problems leading to better learning in CBR systems. In classical CBR techniques, only the problem-solutions were stored in the case base (Aamodt and Plaza, 1994).

4.1.2 Approaches in CBR

(40)

39 remanufacturing have used structural CBR approach to describe the cases(eg. Jiang et al., 2016; Zhou et al., 2014; Veerakamolmal and Gupta, 2002).

Unlike structural CBR in a pure textual CBR approach, all the cases are available in elaborate textual format. They are retrieved using information retrieval techniques with some synonym and semantic ontology described in the system (Weber et al., 2005). Synonym ontology is utilized so as to overcome the synonym problem while retrieving relevant cases. The disadvantage of textual CBR approach is that, with large number of cases in a wide range of domain, the performance of textual CBR will not be better than ordinary information retrieval techniques (Landau, 2010). Since in CBR, the retrieval phase plays a pivotal role in the problem-solving cycle, Bergmann and Schaaf, (2003) suggests incorporating domain ontology in textual CBR systems in order to improve the performance of the CBR system. Other technique include converting knowledge in textual format to structural format using several text mining tools and techniques. Techniques such as text clustering “automatically groups documents by similarity and produces document clusters sometimes described by the words that best represent them”(Landau, 2010, p.64). However, the main disadvantage of this technique is that clustering is fully based on word frequency which might lead to clustering documents based on irrelevant attributes.

4.2 CBR Based Document Management System

As the firm in this research has past remanufacturing experiences stored in centrally accessible DMS, the following section will detail about how CBR techniques can be incorporated in Document Management System (DMS).This is in line with DSR guidelines which prescribe that contextual factors of the application domain must be taken into consideration within which the construction of the artefact will be influenced by the opportunities and constraints of the application domain(Sein et al.,2011,p.38; Hevner and Chatterjee, 2010).

(41)

40 “systems are typically not integrated into an organizations decision-making process” and the systems inability to identify the process or the preconditions to which knowledge applies. Newell, et.al (2006) points out KMS often fail to achieve the stated objective of transferring knowledge across projects due to the inability of the system to retrieve contextual knowledge. This could be because in typical KMS, “users can retrieve necessary information only from the title and description” rather than the actual content of the documents (Cha et.al.,2015, p.203). Even though some systems allow retrieval from automatically produced indices of textual content; more often it fails to retrieve based on the users’ informational context. This is because the automatic search indexing7_{used in typical document retrieval} system assumes that the “best indexing terms are those that occur with high frequency in a document relative to their occurrence in other documents in the collection” (Liddy, 2005, p.9). In addition, DMS typically use vector space model retrieval system which are based on keyword matching and the frequency of words to find the similarity between the inputted search query and the content/file names of the documents (Liddy, 2005; Staab et al., 2001). Consequently, the retrieval is a ranked list of documents based on the term frequency and not related to the applicability of the documented knowledge.

Implementation of CBR systems to overcome these issues requires extensive restructuring of the knowledge present in DMS into a separate software application. In addition, Göker et al. (2010, p.53) points out that the CBR systems are not a standalone application and will fail if it is not “technically and organizationally integrated with the operating environment”. To overcome the aforementioned challenges, in this research, CBR techniques are adopted in DMS. As the core idea behind CBR is efficient knowledge retrieval (Haque et al., 2000), implementation of CBR techniques in DMS will aid retrieval of relevant information based on the applicability of the knowledge. CBR techniques are increasingly used in information retrieval and has been proved to be a powerful and flexible technology as it combines context-based searches and flexible indexing mechanism (Bouhana et al., 2011; Watson and Watson,1998). Due to the strong correlation between CBR cycles and knowledge management cycles, CBR techniques are widely used to guide the design of KM systems and

7_{Search engine indexing collects, parses, and stores data for use by the search engine in the future}

(42)

41 therefore act as the foundational framework for KMS (eg. Weber and Aha, 2003). The association is such that CBR techniques serves as the intelligent component which allows the customization of experience management systems according to organizational needs (Althoff and Weber,2005). The following section describes the detailed design process involved in creating a CBR based Document Management System.

Project-based firms have project experiences stored as documents in detailed textual format. This has certain implications while selecting the approaches taken for the CBR framework. Structural CBR approach cannot be successfully applied since it involves arranging the vast textual contents in tables or in an object-oriented manner which would result in insufficient representation of project experiences (Bergmann and Schaaf, 2003). Whereas; an adoption of pure textual CBR approach would impair efficient retrieval due to large number of documents. Text mining techniques to convert textual contents to structural format would result in grouping contents based on irrelevant attributes (Landau, 2010). Due to the impracticability of conversion an elaborate list of documents to structural format; a semi-structured CBR approach will be utilized in the DMS. This approach involves representing the textual contents of the documents as the “solution part” and the applicable context in which the knowledge should be applied as the “problem description part”. The applicable context i.e. “the problem description part” can be described for each document using the concept of metadata. Metadata is information about data and are search engine parsable(W3School,2017) which allows “identification and easy retrieval of stored documents” (Khan et al., 2015,p. 404). Encoding metadata based on “the problem description part” allows retrieval based on applicability rather than keyword matching.

(43)

42 “problem description part” and the contents of the document acts as the “solution part”.Even though user-defined metadata will be added to the crawled properties8_{during a search crawl,} it won't reflect in a search index. Therefore each metadata in the crawled properties has to be mapped to the corresponding managed properties which would result user-defined metadata reflect in the search index. Managed properties are lists of data containing contents and metadata from which search index crawls to find relevant documents based on the refined search queries (Microsoft,2016).Once mapped; improved retrieval of relevant documents can be achieved through exploiting the metadata annotated documents by the search functionality existing with DMS.Since “problem description part ” annotated in documents have multiple attributes in order to describe the applicable context, parametric search must be employed in DMS.This can be accomplished in the incumbent system, by mapping the fields in the parametric search to the managed properties of the metadata. This effectively allows the search system to compare the similarity between the corresponding metadata and the search attributes entered by the user.Moreover, a set of specific fields will guide and restrict users forming focussed representations that will make it easier to bring in context from the user and will help the system to identify a documents usefullness and applicability (Weber,2007).Even though parametric search allows the user to describe the search attribute to a detail extend, it could slow down the DMS retrieval performance(Microsoft 2010). The figure below shows a screenshot of parametric search implemented in Sharepoint.

Figure 6 Parametric Search in SharePoint Source (Microsoft 2010)

As in structural CBR, a domain ontology can be pre-defined for describing the metadata of the document. This allows a uniform vocabulary for describing “problem description part” and hence will help DMS search engine to interpret the description of the

8_{A crawled property is content and metadata that is extracted from documents during search}

(44)

43 problem by overcoming the synonym problem (Potes Ruiz et al., 2013). This would facilitate retrieval based on “concept based contents rather than keyword-based contents” (Han and Park, 2009,p.7442).This approach was utilized by Staab et al., (2001) to retrieve knowledge items based on metadata which was based on the ontology of the subject domain. In the incumbent system, consistent encoding of metadata according to the domain ontology can be accomplished by use of managed term sets. Managed term sets are a predefined which allows only a list of specified terms to be encoded as metadata. These term sets can be then classified into taxonomies which let term sets to be grouped into hierarchy(Microsoft,2013).Taxonomies can organize metadata into hierarchies, stores information at appropriate levels of generality (Kapetanios and Kramer, 1995).Thus, taxonomies help in defining the relationship between each term sets. Taxonomy being a subset of ontology can capture the parent-child relationships in a particular domain(Han and Park, 2009).The below figure 5 shows metadata annotation of a document and the product taxonomy formed using term sets.

Figure 7 Annotation of meta using term set hierarchy Adopted from Microsoft Office Support

(45)

44 conceptualization of a domain which can be shared among the employees to facilitate knowledge access, sharing and reuse (Staab et al., 2001). This notion of ontology as a shared concept within an organization is well established and has been used widely in knowledge management, knowledge acquisition, information retrieval and extraction.In knowledge management, ontology can be used as a basis for “context-aware content description of knowledge” (Han and Park, 2009,p.7442).In information retrieval, search terms based on ontology can prevent the synonym problem which may cause mismatching of similar terms (Lau et al., 2009) and can facilitate multi-layered and networked knowledge retrievals(Han and Park, 2009).However, the creation of ontology require heavy engineering effort and man-hours to identify the domain instances belonging to a concept and their semantic relationships (Zidi et.al,2014).In addition, ontology will have to reflect the changing environment of the firm(Staab et al.,2001) which consecutively would affect the encoded metadata.

Decision Support Framework for Remanufacturing Process Planning: A Design Science Research

MASTER THESIS