Capturing and Reusing Knowledge

Embed Size (px)

Citation preview

  • 8/20/2019 Capturing and Reusing Knowledge

    1/20

    Capturing and reusing knowledge in engineering change

    management: A case of automobile development

    Hong Joo Lee   & Hyung Jun Ahn   & Jong Woo Kim   &

    Sung Joo Park 

    Received: 27 June 2006 /Accepted: 5 September 2006 / Published online: 5 December 2006# Springer Science + Business Media, LLC 2006

    Abstract   The development of complex products, such as

    automobiles, involves engineering changes that frequentlyrequire redesigning or altering the products. Although it has

     been found that efficient management of knowledge and

    collaboration in engineering changes is crucial for the

    success of new product development, extant systems for 

    engineering changes focus mainly on storing documents

    related to the engineering changes or simply automating the

    approval processes, while the knowledge that is generated

    from collaboration and decision-making processes may not 

     be captured and managed easily. This consequently limits

    the use of the systems by the participants in engineering

    change processes. This paper describes a model for 

    knowledge management and collaboration in engineering

    change processes, and based on the model, builds a

     prototype system that demonstrates the model’s strengths.

    We studied a major Korean automobile company to analyze

    the automobile industry’s unique requirements regarding

    engineering changes. We also developed domain ontologies

    from the case to facilitate knowledge sharing in the design process. For achieving efficient retrieval and reuse of past 

    engineering changes, we used a case-based reasoning

    (CBR) with a concept-based similarity measure.

    Keywords  Automobile development . Case-based

    reasoning . Engineering change management . Knowledge

    capturing . Knowledge reuse . Semantic web

    1 Introduction

    Engineering changes involve the modification of products

    and components that occur after the product design is

    released (Clark & Fujimoto,   1991; Huang, Yee, & Mak,

    2003; Terwiesch & Loch,   1999). Development processes

    for complex products usually involve many engineering

    changes, which mostly reflect technological advances,

    resolve defects in the design, and improve the overall

    quality of the products (Adler & Clark,   1991; Pikosz &

    Malmqvist,   1998; Terwiesch & Loch,   1999). Engineering

    changes are considered inevitable, especially for complex

     products, because product development usually takes a long

    time and involves collaboration among designers and

    engineers who are often geographically distributed (Huang

    et al., 2003).

    In order to reduce the development cost and time

    required to produce a new product, many researchers and

     practitioners are interested in the efficient management of 

    knowledge and collaboration in engineering change pro-

    cesses. Studies have found that engineering changes for 

    complex products are often performed by heterogeneous

    teams in distributed environments, where knowledge-

    Inf Syst Front (2006) 8:375 – 394

    DOI 10.1007/s10796-006-9009-0

    H. J. Lee (*)

    Center for Coordination Sciences, Sloan School of Management,

    Massachusetts Institute of Technology, NE20-336,

    3 Cambridge Center, Cambridge, MA 02139, USA

    e-mail: [email protected]

    H. J. Ahn

    Management Systems, Waikato Management School,University of Waikato, Gate 1 Knighton Road,

    Private Bag 3105, Hamilton, New Zealand

    J. W. Kim

    School of Business, Hanyang University, 17, Haengdang-dong,

    Seongdong-gu, Seoul 133-791, South Korea

    S. J. Park 

    Graduate School of Management,

    Korea Advanced Institute of Science and Technology,

    207-43, Cheongryangri-dong, Dongdaemun-gu,

    Seoul 130-722, South Korea

  • 8/20/2019 Capturing and Reusing Knowledge

    2/20

    intensive collaboration and communication for various

     problem-solving and decision-making are essential (Alavi

    & Leidner, 2001; Grant,  1996). However, existing systems

    that support engineering change management (ECM) are

    mostly limited to the simple issuing and approval of 

    engineering change orders (ECOs) through workflow

    management systems or storage and keyword-based re-

    trieval of engineering change documents, which makes it difficult to capture and reuse the informal, unstructured

    knowledge inherent in engineering change processes. For 

    this reason, a huge part of the valuable knowledge that has

     been generated during the past collaborations, decision-

    making, and from the contextual relationships between

    different types of knowledge can be lost, which often

    results in limited use of the systems. Since much of the

    engineering change knowledge are tacit and unstructured, it 

    is very inefficient to use the existing support systems for 

    sharing the knowledge among the designers and for solving

     problems (Ahn, Lee, Cho, & Park,   2005; Grant,   1997;

     Nonaka & Konno, 1999).The purpose of this paper is to develop a model and

     prototype support system for ECM to facilitate the accumu-

    lation and reuse of the knowledge generated in collaborative

    engineering change processes. We investigated the engi-

    neering change processes of a major Korean automobile

    company to determine the requirements of ECM for a

    specific industry. Then, a collaboration model was designed

    to support the geographically distributed designers and

    decision-makers involved in the engineering change pro-

    cesses. The collaboration model is also used as a knowledge

    annotation schema along with a Semantic Web (Berners-

    Lee, Hendler, & Lassila, 2001) language, so that knowledge

    items in ECM can accumulate naturally through the

     process, and contextual knowledge can also be captured

    easily. Domain ontologies are also developed to represent 

    the knowledge context. To efficiently retrieve and reuse

     past cases of engineering change, a case-based reasoning

    (CBR) technique is used with the concept-based similarity

    measure (Resnik,   1999). A prototype ECM system is

    implemented as a demonstration of the proposed model.

    The remainder of this paper is organized as follows.

    Section  2   reviews the literature on ECM issues. Section   3

    describes a case study of the ECM practices in a Korean

    automobile company. Section 4 describes the derived modeland examples of knowledge accumulation and retrieval.

    Section   5   presents the prototype system, which is called

    CECM (Collaborative environment for Engineering Change

    Management). Section 6   discusses the implications of this

    study and concludes with suggestions for further research.

    2 Research background

    2.1 Engineering change management 

    Engineering changes are not always to the detriment of adevelopment project, as many cost savings or performance

    improvements are brought into the project in the form of 

    engineering changes (Clark & Fujimoto,   1991; Clark &

    Wheelright,   1993; Terwiesch & Loch,   1999). Table   1

    summarizes the major causes of engineering changes. Al-

    though there are unnecessary changes that should be avoided,

    many of the engineering changes are actually beneficial.

    Therefore, it is neither desirable nor realistic to focus one ’s

    efforts on simply eliminating the engineering changes (Clark 

    & Fujimoto,   1991). Consequently, it is more important to

    manage the engineering change processes efficiently to

    reduce the time and cost of product development.

    The formal process of an engineering change usually

    consists of four stages (Huang, Yee, & Mak,   2001;  2003;

    Pikosz & Malmqvist,   1998; Terwiesch & Loch,   1999):

    initiating an engineering change request (ECR), evaluating

    Table 1   Causes of engineering changes

    Causes of engineering changes Descriptions

    Careless mistakes Corrections of errors on a document (Clark & Fujimoto, 1991; Clark & Wheelright,  1993;

    Pikosz & Malmqvist,  1998)

    Poor communication Faults in the interpretation of customer demands into technical requirements (Pikosz & Malmqvist, 1998)

    Changes in the customer specification (Pikosz & Malmqvist,  1998)

    Changes in the manufacturing process or situations

    Snowballing change Change of a part depending on altered function or production requirements (Huang & Mak, 1998;

    Pikosz & Malmqvist,  1998; Wright, 1997)

    Organizational, technological and operational changes (Karvounarakis et al.,  2003)

    Cost savings Cost savings (Clark & Fujimoto, 1991; Clark & Wheelright,  1993)

    Change, replacement, withdrawal, and introduction of a part (Pikosz & Malmqvist,  1998)

    Ease of manufacturing Difficulties in parts fabrication or assembly (Pikosz & Malmqvist, 1998)

    Product performance improvement Weakness in the product identified during prototype testing (Pikosz & Malmqvist, 1998)

    Quality problems with some subsystems or component (Pikosz & Malmqvist,  1998)

    Development for future product revisions (Pikosz & Malmqvist,  1998)

    376 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    3/20

    the ECR, issuing engineering change orders (ECOs) to

    relevant participants, and storing and analyzing the ECOs

    for management purposes (Fig.  1).

    Based on the four-stage model, the characteristics of 

    typical engineering change management processes and

    associated problems can be identified as shown in Table  2

    (Adler & Clark,  1991; Clark & Fujimoto,  1991; Huang &

    Mak, 1998, Huang et al., 2001; 2003; Pikosz & Malmqvist,

    1998; Terwiesch & Loch,  1999).

    &   Complexity: The products and processes involved in

    engineering changes usually involve a high level of 

    complexity. Complexity in processes requires iterative

    communication and collaboration to adjust the engineering

    changes. Complexity in products incurs snowballing effect 

    when a change in a single part is propagated to numerous

    related parts or products, which often contributes to long

    ECO lead times (Terwiesch & Loch, 1999).

    &   Virtual collaboration: Today, engineering changes for a

    complex product like an automobile usually involve the

    collaboration of virtual design teams, which operate in a

    geographically distributed or global environment.

    &   Non-routine, knowledge-intensive tasks: Many of the tasks

    involved in engineering change processes are non-routine

    and require problem solving by heterogeneous groups of 

     people with high levels of expertise. This makes knowl-

    edge sharing and transfer between team members critical.

    Hence, support systems for ECM should be designed to

    reduce the complexity of the process and to facilitate virtualcollaborations and knowledge management.

    2.2 Computer-aided ECM systems

    Early studies of computer tools for ECM have focused on

    stand-alone computer-aided systems for storing and retrieving

    engineering change documents (Huang et al.,  2001; Wright,

    1997) to replace paper-based ECM. In some companies, the

    engineering changes are managed by configuration manage-

    ment (CM) systems which focus on how products are

    configured and changed (Huang & Mak,  1998).

    Recently, ECM systems based on workflow systems

    have been proposed (Huang & Mak,   1998; Huang et al.,

    2001) that provide a wide set of functionalities for 

    supporting the entire engineering change process. They

    allow employees to issue and approve engineering change

    requests, track the change status, and manage the related

    documents. Although the workflow-based systems manage

    documents efficiently, they are still limited in their ability to

    capture and reuse the knowledge involved in engineering

    changes. Important knowledge resources, such as context 

    information on knowledge items, collaborative experiences,

    and decision-making processes, are not contained in ECM

    systems because the systems cannot incorporate collabora-

    tive change processes. Engineering change teams currently

    depend heavily on off-line collaboration without codifying

    Table 2   Characteristics and problems in the engineering change process

    Characteristics of engineering

    change processes

    Problems

    Complex Process Extensive document management (Clark & Fujimoto, 1991; Pikosz & Malmqvist,  1998)

    Complex approval process (Terwiesch & Loch,  1999)

    Long learning time for new employees and consultants (Pikosz & Malmqvist,  1998)

    Product complexity Induce a series of downstream changes — snowballing changes (Huang & Mak, 1998; Pikosz &

    Malmqvist, 1998; Terwiesch & Loch,  1999)

    Couplings between the component or development activities (Terwiesch & Loch,  1999)

    Ad hoc activities Limit capacity of an individual engineer (Terwiesch & Loch, 1999)

    A significant time loss from   “diving into the project again”   (Loch & Terwiesch, 1996;

    Terwiesch & Loch, 1999)

    Distributed environment Distributed members (Huang et al.,  2003)

    Cross functional and multi-disciplinary teamwork (Huang et al.,  2001)

    Alternative solutions are evaluated to satisfy everyone (Pikosz & Malmqvist,  1998)

    Conflict between departments and cultural differences (Terwiesch & Loch,  1999)

    Knowledge intensiveness Creativity and innovation are highly required (Adler & Clark, 1991)

    High reuse the information is necessary (Pikosz & Malmqvist,  1998)

    ECRForm

    ECREvaluation

    Form

    ECOForm

    Problem

    Detection

    Identifying

    Problem Scope

    Generating

    Alternatives

    Approval

    Process

    Engineering

    Change Log

    ReferenceReference

    Fig. 1   The process of engineering change management 

    Inf Syst Front (2006) 8:375 – 394 377

  • 8/20/2019 Capturing and Reusing Knowledge

    4/20

    the knowledge explicitly. Consequently, knowledge reuse is

     poor because of the insufficient management of past 

    experience and is limited to searching by keywords or 

    reference numbers and hence browsing categories.

    In addition to workflow, other approaches exist for 

    implementing ECM systems. There are several knowl-

    edge management (KM) systems for product development 

    that are intended to capture process knowledge and toshare product data (May, Carter, & Joyner,   2000; May &

    Carter,   2001; Monplaisir,   1999; Numata & Taura,   1996;

    Ramesh & Tiwana, 1999; Yoo & Kim, 2002). Ramesh and

    Tiwana focused on capturing process knowledge for 

    collaborative product-development tasks, and suggested a

    model that represented concepts in product development 

    along with a concept map describing dependencies among

    the concepts (Ramesh & Tiwana, 1999), which is similar to

    the issue-based information system (IBIS) model (Conklin

    & Begeman, 1988). Computer supported cooperative work 

    (CSCW) has also been applied to product development 

    (May et al., 2000; May & Carter, 2001; Monplaisir, 1999).It has been shown that the productivity of engineering

    design teams can be enhanced further when the CSCW

    functions are augmented with integrated process develop-

    ment architecture (May et al.,   2000; May & Carter,  2001;

    Monplaisir,   1999). Numata and Taura (1996) emphasized

    communication among engineers participating in product 

    development and suggested a knowledge network for 

    enhancing the convenient transfer of knowledge among

    engineers (Numata & Taura,  1996). These systems demon-

    strate the use of typical KM functions for ECM, but they

    are still limited in organizing and sharing product knowl-

    edge manually so that contextual knowledge is usually

    difficult to capture.

    3 Case study

    3.1 Overview

    As a case study, we investigated the new product development 

     projects of a major Korean automobile company. Data were

    collected from multiple sources, including the following:

    &   Semi-structured interviews with staff members responsiblefor engineering, operation management, cost management,

    and development process management.

    &   A collection of articles on engineering changes and new

     product development that were submitted as a result of the

    host company’s internal training program.

    &   Data analysis from a Web-based information system that 

    the company uses to keep track of ECOs.

    The company’s product development teams are distrib-

    uted globally in 14 plants and six R&D centers. Product 

    engineering activities are closely related to concept gener-

    ation, product planning, and process engineering activities,

    and need intensive communication with suppliers. Product 

    development teams use a workflow system to handle

    engineering changes in which a simple workflow instance

    is defined as initiating ECRs, issuing and noticing ECOs,

    and approving ECOs. The company’s Web-based intranet 

    system includes e-mail, community discussion boards, anda document management system.

    3.2 The process of engineering change

    Figure   2   describes the steps needed before an engineering

    change of an automobile component is handled successful-

    ly. When a problem is detected in a certain process, a

    request is made for the engineering team or the engineer in

    charge of that automobile component to review the

     problem. Through discussions among engineers and team

    members who detected the problem, they decide whether toinitiate an ECR. When a decision is made to initiate an

    ECR, the team completes the ECR form using the workflow

    system. This form contains information on the product and

    automobile components that need to be changed, and the

    reasons for and a description of the ECR. After the

    engineering team reviews the ECR, they decide to accept 

    or reject it. The status of an ECR can be tracked and

    monitored by the person who made the request. The

    assigned engineer identifies the causes of the problem and

    the scope of the engineering change by communicating

    with related engineers. To generate alternative solutions to

    the engineering change, they usually rely on their past experience and knowledge. In addition, they try to find

    similar situations from the intranet system, which stores the

     problem and engineering change reports from past projects.

    Then, the engineer in charge issues an ECO that contains

    the information from the ECR; changes in weights and

     production costs, the time when the engineering changes

    will become effective, and the plans to handle obsolete

     parts. If there is a change in the unit cost of a part, the ECO

    has to be approved by the cost management department 

    simultaneously. After administrative approval, the ECOs

    are sent to the technology management team to validate the

    consistency in the product design and to check for possibleerrors and inconsistencies. Then, the ECOs and associated

     product data are released and stored in the intranet system

    of the company.

    3.3 Knowledge management practices

    To derive the desired features of ECM, the ECM process of 

    the automobile company is reviewed and analyzed from a

    knowledge-management perspective, emphasizing the four 

    378 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    5/20

    stages of the knowledge process: knowledge creation, storage/ 

    retrieval, transfer, and application (Alavi & Leidner, 2001):

    &   Knowledge Creation: Despite the knowledge-intensive

    nature of the engineering change process, there is no

    formal consideration of knowledge management related

    to ECM in the automobile company. An engineering

    change process is regarded as merely an instance of the

    company’s workflow, and the resulting knowledge is

    embedded in separate documents. The current system

    fails to capture the variety of knowledge associated with

    engineering changes. In addition, the knowledge created

    in off-line collaboration is not incorporated in the current 

    workflow system because the system does not provide a

    way to link informal, unstructured knowledge with the

    structured knowledge in workflows. The rigidity of the

    document form also limits the generation and accumu-

    lation of knowledge.

    &   Knowledge Storage/Retrieval: All ECRs and ECOs are

    stored automatically in the intranet system after the

    workflow is finished. The engineering changes are orga-

    nized by product category, component modules, and part 

    information. The intranet system does not provide suffi-

    cient support of the contextual knowledge which can be

    used to navigate for the relevant knowledge. The valuable

    links between a specific engineering change and related

     parts, products, problems, and solutions are not preserved

    in the system. Past engineering changes are retrieved using

    a simple keyword-based search and browsing predefined

    categories of engineering changes.

    &   Knowledge Transfer: According to a company report on

    new automobile development projects, one-third of all

    engineering change problems are not new. As a universal

    characteristic of automobile companies, product engineer-ing departments are highly compartmentalized according to

    the models and functions of the automobiles. Knowledge

    related to the same problems in different components or 

    the same component in different models cannot be

    transferred easily.

    &   Knowledge Application: Since ECM systems merely store

    minimal knowledge of engineering changes, valuable

    knowledge, such as difficulties in making changes and

    important decision-making issues, is lost. The engineers

    depend mainly on their tacit knowledge of past experience

    and off-line communications to solve the engineering

    changes at hand.

    3.4 Findings from the case study

    The engineering change process of an automobile company

    is complex; it handles various types of knowledge, and

    requires collaboration among distributed engineers. At the

    automobile company of the case study, the engineers in the

    engineering department usually regarded engineering

    changes raised in the latter part of the development as their 

    Fig. 2  The process of engineering change

    Inf Syst Front (2006) 8:375 – 394 379

  • 8/20/2019 Capturing and Reusing Knowledge

    6/20

    fault. Hence, they often tried to handle the changes quickly

    using off-line collaboration without using excessive, time-

    consuming paperwork. This made it difficult to capture the

    tacit knowledge embedded in the collaboration and pre-

    venting the use of similar cases in later stage. Even for the

    engineering cases preserved in the current system, the

    stored knowledge are not sufficient to capture the rich

    relationships between design knowledge items. The system provided access to the database via simple keyword-based

    queries and category-based browsing. These problems have

    limited the accumulation of knowledge which led in turn to

    ineffective and inefficient use of the system. Although the

     problems found in the case study are specific to the target 

    organization, we reasonably postulate that many of the

     problems stem from the inherent complexity of general

    engineering change processes.

    4 The collaboration model for engineering

    change management

    The objective of this research is to design a collaborative

    environment for ECM (CECM) to facilitate the accumula-

    tion and use of knowledge amassed through the engineering

    change processes of the target organization as shown in

    Fig. 3. The key points in the approach are as follows.

    First, a model is developed that captures the nature of the

    collaboration in engineering processes. The model tries to

    support both routine and non-routine collaboration to

    overcome the limitations of many workflow-based systems

    that only save form-based documents and use them in

    routine workflow.

    Second, the figure shows that the collaboration model is

    used for knowledge management functions of CECM as

    well. The collaboration model plays an important role in

    capturing both general and contextual knowledge naturally,

    along with collaborative processes. In other words, it not 

    only provides a basis for collaboration support but also is

    used as a schema for knowledge representation for the

    CECM. Aligning knowledge management with collabo-

    ration can be a very powerful method of supporting

    knowledge-intensive work in a distributed collaborative

    environment.

    Third, for the knowledge management functions, we

    develop ontologies, design a case-based reasoning mecha-

    nism for knowledge retrieval in CECM, and provide

    resource description framework (RDF)-based Semantic

    Web representations of engineering change knowledge.The ontologies provide common representations of various

    types of knowledge used for engineering processes, which

    are used to create Semantic Web pages for knowledge

    sharing. Combining the case-based reasoning mechanism

    with the ontologies is useful to retrieve engineering change

    cases in practice.

    4.1 Design of the collaboration model for engineering

    change management 

    Since the engineering change process involves many

    knowledge-intensive tasks performed in a distributed

    environment, we adopted the knowledge context model

    (KC-V) as a base model. The KC-V model was originally

    developed to capture and to utilize contextual knowledge in

    virtual collaboration environments. The KC-V identifies

    three key perspectives of the modeling context in virtual

    environments:  organization,   activity, and  context  (Abecker,

    Bernardi, Hinkelmann, Kühn, & Sintek, 2000; Jarvenpaa &

    Leidner, 1999; Maus,  2001; Wong & Burton,  2000). It has

     been used to build a knowledge and collaboration manage-

    ment system for virtual work called the Virtual Workgroup

    Support System (VWSS) (Ahn et al.,   2005). The most 

    distinguishing feature of KC-V and VWSS is that they

    integrate collaboration with knowledge management, when

    knowledge is captured naturally along with activities. The

    coordination and execution of activities naturally contribute

    to the accumulation of knowledge, and in doing so, the

    knowledge context is effectively preserved based on the

    context structure of the KC-V. Hence, we developed our 

    collaboration model by refining the KC-V and reflecting

    unique characteristics of the problem domain as shown in

    Fig. 4.

    The three perspectives of the KC-V are applied to theengineering change process as follows. The organizational

     perspective explicitly defines and supports the life cycle of 

    virtual teams that involves organizing the virtual engineer-

    ing change teams, collaborating until the engineering

    change problems are solved, and disbanding the teams

    after the goals are reached. The activity perspective

    emphasizes the alignment of knowledge management and

    collaborative activities. It tries to link the outputs of routine

    and non-routine activities of engineering change manage-

    ment to a knowledge repository and preserve the rich

    CollaborationModel

    Collaborative Environment

    for ECM (CECM)

    Collaboration

    Support:Routine and Non-routine

    collaboration

    Knowledge Management:Ontology, Case-Based

    Reasoning, andSemantic Web

    Fig. 3  Overview of the approach

    380 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    7/20

    relationships among knowledge items so that they can be

    used as contextual knowledge. Instead of merely storing

    engineering change forms in a document base, for example,

    it can turn discussion messages into knowledge items;

    answering various workflow requests or ad hoc non-routine

    conversation messages naturally contributes to the genera-

    tion and accumulation of knowledge; setting, coordinating,

    and reaching milestones of activities are also associatedwith knowledge creation so that additional efforts at 

    collecting knowledge can be minimized. Thus, the pro-

     posed system based on the KC-V is neither a collaboration

    system nor a knowledge management system but a

    combination of both that preserves the rich knowledge

    context. The third component, the context perspective,

    simply helps to organize the knowledge items by linking

    entities of the organizational perspective and the activity

     perspective (Ahn et al.,  2005).

    4.1.1 Activity and process

    A process can have hierarchically organized activities as

    shown in Fig.   4. It also has the life-cycle status of an

    engineering change process as an attribute. After complet-

    ing an engineering change, the entire set of knowledge

    items and the structure of the activities used in the process

    can be stored in the knowledge repository, along with the

    knowledge context. Placed at the center of the model, the

    activity is crucial for integrating knowledge items with a

    knowledge context. Knowledge items are associated with

    activities via the knowledge context entity, and other 

    context entities, such as actors, milestones, and conversa-

    tions, can also be associated with the activity entity.

    4.1.2 Conversation

    A conversation is a set of structured messages based on

    speech-act theory for supporting cooperative processes

    (Agostini, De Michelis, & Grasso,   1997; Delteil, Faron-

    Zucker, & Dieng,  2001). During a conversation process,

    knowledge items, such as discussions and documents, are

    created and the status of an associated activity or document is updated.   ‘Conversation’   is a simplification of the

    ‘Coordination’   construct in the KC-V model. This change

    was made because the coordination construct in the KC-V

    model was aimed at very broad types of coordination that 

    are not necessary for the engineering change process.

    4.1.3 Two types of knowledge item: document 

    and discussion

    The collaboration model supports two types of knowledge

    item: document and discussion entities. The document entity

    represents formal, structured documents and forms, such asECRs, ECOs, and interim or final reports of an engineering

    change activity. In contrast, the discussion entity was de-

    signed to broadly represent unstructured communications

     between actors, and is associated with activity, document,

    and conversation entities.

    4.1.4 Knowledge context 

    The knowledge context entity defines the context of a

    knowledge item explicitly, as depicted in the Fig.  4. In this

    way, each knowledge item, such as documents and discus-

    sions, is explicitly associated with its creation context, which

    ActivityProcess

    ConversationActor DocumentDiscussion

    KnowledgeItem

    Role

    1

    Milestone

    1

    Concept

    0..*

    .

    1

    .

    1

    .

    0

    .1

    Ontology

    1 p ar 

     t  i   c i   p a t   e s 

    pertains to

     o

    wn s 

     b  el   on g s  t   o

    produces

    initiates

      p e  r  f o

      r  m  s      h    a    s

    is related with

    coordinates

    refers

    h     o    

    l     d      s    

    includes

    containsKnowledge

    Contex t

    is super of

    possesses

    ScheduleEvent

    preserves

    Fig. 4   The collaboration model

    Inf Syst Front (2006) 8:375 – 394 381

  • 8/20/2019 Capturing and Reusing Knowledge

    8/20

    is centered on the activity entity within which the knowledge

    item was created. Through this explicitly defined knowledge

    context entity, the system can provide comprehensive links

    to entities that are helpful for understanding individual

    knowledge items, such as related activities, actors, conver-

    sations, and milestones.

    4.1.5 Ontology

    The   ‘ontology’   construct in the collaboration model is

    derived from the  ‘domain’ construct of the KC-V and is used

    to retrieve past knowledge items reflecting the complex

    multidimensional characteristics of the automobile industry.

    For an automobile development project, the following

    domain ontologies were identified as key dimensions

    (Golebiowska, Dieng-Kuntz, Corby, & Mousseau, 2001):

    &   Product ontology: a hierarchy of product segmentations

    and their manifestations, such as passenger cars, recreation

    vehicles, and commercial vehicles.&   Component ontology: a hierarchy of components, modules,

    and functions in a vehicle, such as engine, cockpit, and axle.

    &   Problem ontology: a collection of problem types that 

    describes the causes of problems and reasons for engineer-

    ing changes, such as product safety and manufacturing

    difficulties.

    &   Solution ontology: a hierarchy of solution types in

    engineering changes, such as product form modification

    and assembly hole relocation.

    &   Process ontology: the structure of a new product development 

     process and its hierarchy of activities in a project, such as

     planning, styling, or pilot production stages.

    Figure 7  shows examples of these ontologies.

    4.2 Knowledge representation with the collaboration

    model and ontologies

    The collaboration model can be a schema of knowledge

    annotation used to define knowledge structure and validate

    knowledge in a knowledge management application. All of 

    the constructs in the collaboration model are used to

    annotate the knowledge items, along with their context 

    information. Resource description framework schema(RDFS) is a schema-specification language developed for 

    representing the RDF statements (RDF, 1999; RDFS, 2000)

    used for the Semantic Web (Berners-Lee et al.,   2001;

    Decker, Mitra, & Melnik,   2000; RDF,   1999). An RDF

    annotation consists of a set of statements, each one

    specifying the property of a resource (Decker et al.,  2000;

    De Michelis, & Grasso,   1994; Melnik,   2000; Melink &

    Decker,   2000; RDF,   1999). Based on the RDFS descrip-

    tions, instances of domain ontologies and the entities of the

    collaboration model are generated. The RDF instances of 

    the domain ontologies are also used in defining a case.

    Consequently, the knowledge items in engineering change

     processes are annotated using the RDF definitions for 

    outputs such as ECOs, modification results, and problem-

    solving reports.

    The Semantic Web-based knowledge representation

    combined with the collaboration model and domain

    ontologies has the following advantages:

    (1)   Accumulation of knowledge context: As shown in Fig. 5,

    a knowledge annotation includes contextual information

    of knowledge items, such as related activities, milestones,

    and actors. Users can navigate to other relevant knowl-

    edge items through the links provided by the knowledge

    context of a given knowledge item.

    (2)   System-independent representation of knowledge: Se-

    mantic Web technology such as RDF enables us to enrich

    knowledge by annotating or giving its meaning with tags

    defined in the ontologies. The structure and meaning of 

    RDF statements can be identified by other systems withRDFS and associated ontologies. For example, using RDF

     pages, it becomes easier for other systems, such as a

     product data management (PDM) system, to identify the

    meaning of engineering changes, such as related parts, the

    source of problems, and problem-solving methods.

    4.3 Knowledge reuse in engineering change management 

    Ontology-driven knowledge management answers queries

    using logical deduction based on ontologies (Broekstra,

    Kampman, Harmelen, & Sesame,   2003; Karvounarakis

    et al.,   2003), but the strictness of deductive reasoning has

     been recognized as one of the major obstacles in ill-

    structured and complex problems (Bergmann & Schaaf,

    2003).

    Case-based reasoning (CBR) can be used to relax the

    strictness by allowing the retrieval of knowledge items that 

    are close to queries, but not exact matches (Bergmann &

    Schaaf,   2003; Kolodner,   1993; Lorenzo & Perez,   1997).

    However, there are difficulties in applying CBR to ECM.

    First, an engineering change case (e.g., ECR or ECO) is

    defined as an aggregate of the related components, processes, problems, products, and solutions that does not 

    allow easy quantification for developing numerical distance

    measures. For example, to define the numerical distance

     between a 2.5 DOHC gasoline engine and a 2.0 diesel

    engine, one may have to rely on experts’   subjective

    opinions. Furthermore, the manually assigned weights and

    similarity measures can easily be outdated because of the

    introduction of new products or new components. In

    addition, the large number of different parts and compo-

    nents makes the experts’   subjective weighting almost 

    382 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    9/20

    impossible (Kolodner,   1993). In order to address this

     problem, the information-based measure of Resnik (1999),

    which measures concept similarities automatically, is adopted

    (Resnik,  1999). In addition, since an engineering change

    case is defined using several different ontologies, proper weights should be assigned to the ontologies depending on

    their relative importance in retrieving engineering change

    cases. For example, we need to determine the relative

    importance of the   ‘ process aspect ’   of an engineering

    change case compared to the   ‘component aspect.’   To

    determine the weights, the analytic hierarchy process

    (AHP) was used (see Appendix   A) (Chen & Huang,

    2001; Saaty,  1980). A set of pair-wise comparisons of theontologies was collected from the experts in the case

    company using questionnaires. The result of the AHP

    calculation is presented in Fig. 6.

    0.447

    0.21

    0.035

    0.094

    0.214

    0 0.1 0.2 0.3 0.4 0.5

    Component

    Problem

    Process

    Product

    Solution

    Normalized Weights

    Fig. 6   Weights of the five

    ontologies

    Engineering Change Case

    Relevant CasesCase information

    -Product-Component

    -Process

    -Problem

    -Solution

    Activity information

    -Activity hierarchy

    -Subordinated items

    Related Document

    -Milestone

    -Preparation material

    Related Discussion

    -Discussed articles

    -Decision making history

    Related Activities & Items

    Case Based Reasoning

    Knowledge

    Context

    Fig. 5   Knowledge context and

    navigation paths

    Inf Syst Front (2006) 8:375 – 394 383

  • 8/20/2019 Capturing and Reusing Knowledge

    10/20

    The final similarity measure between two engineering

    change cases, E1   and E2, is given in formula (1) (see

    Appendix B).

    Sim E 1; E 2ð Þ ¼

    Pi

    wi   f  i   simi   c1i; c2ið ÞP

    i

    wi;   ð1Þ

    where   wi   is the weight of ontology   i   calculated using theAHP,   f  i   is the factor used to reduce the effects of different 

    depths in different ontologies (see Appendix B), and c1i and

    c2i   represent the respective concepts of   E 1   and   E 2   in

    ontology  i.

    As an illustration, consider the situation in Fig.  8. In this

    example, there are two engineering change cases; case 1 is

    an ECR and case 2 is an ECO, and these two cases are

    associated with the ontologies in Fig.  7. Arrows are used to

    indicate comparison points between the two cases; the pro-

    duct concept in case 1 is compared to the product concept 

    in case 2. In addition, the component, problem, and process

    concepts of case 1 are compared to the corresponding con-cepts of case 2. The table in the lower part of Fig.  8   shows

    how each concept in the different ontologies is used to cal-

    culate the aggregate similarity. The third column (Closest 

    Common Parent) shows the closest common parent between

    two concepts in the two cases and the fourth column pro-

    vides the information value of the closest common parent in

    the third column. The fifth column represents compensation

    values for reducing the size effect of ontologies because the

    average value of the information-based measure is propor-tional to the size of a given ontology (see Appendix B). The

    sixth column contains the weights of the ontologies calcu-

    lated using the AHP.

    This CBR method has the following additional advan-

    tages in retrieving engineering change cases. First, diverse

    factors pertaining to engineering changes can be considered

    to find similar cases of past engineering change. In this

     paper, components, processes, problems, products, and solu-

    tions are reflected as the attributes of engineering changes.

    Second, although there can be frequent changes in individual

    items in the ontology, it is unlikely that the relative

    importance of the ontologies will change markedly over time. Therefore, once AHP weights are calculated, they will

    Problem ontology   Component ontology

    p=0.3info = 1.2039

    Assembly

    Screwing

    ResistanceFixing Installation Implementation

    Clipp ing Stapl ing Coupl ing F it ting

    p=0.3info = 1.2039

    p=0.1info = 2.3025

    p=0.05info = 2.9957

    p=0.03info = 3.5065

    p=0.001info = 6.9077

    p=0.049info = 3.0159

    Problemp=1

    info = 0

    Performancep=0.2info = 1.6094

    Component

    Axle Steering

    Suspension  Axle

    Mechanism

    Chassis

    Pedal

    FrontSuspension

    RearSuspension

    p=1info = 0

    p=0.4

    info = 0.9162

    p=0.2info = 1.6094

    p=0.05info = 2.9957

      p=0.1info = 2.3025

    Engine

    p=0.2info = 1.6094

    ClimateControl

    p=0.1info = 2.3025

    Brake

    p=0.5

    info = 0.6931

    Vehicle

    Passenger Car SUV

    Compact Midsize

    Car

    Van

    A3   X7

    p=1info = 0

    p=0.7

    info = 0.3566

    p=0.05info = 2.9957

    p=0.1info = 2.3025

    Bus  Truck

    Fullsize

    Product ontology

    Luxury

    X8X6X5M4

    Process ontology

    Process

    Prototype #1

    Pilot

    p=1

    info = 0

    Engineering   Manufacturing

    p=0.1info = 2.3025

    p=0.2info = 1.6094

    Prototype

    Prototype #2   Assembly Planning Pilot Production

    p=0.05info = 2.9957

    p=0.03info = 3.5065

    Fig. 7   Parts of ontologies

    384 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    11/20

    remain valid for a long time and not require frequent updates.

    Third, the similarity measures can be calculated automati-

    cally from the engineering change case base without human

    intervention. This automatic calculation keeps the cost of 

    expanding the case base for new products and components

    low. Finally, the retrieval mechanism can be flexible and

    accommodate new types of ontologies when users wish to

    include additional dimensions to engineering change cases

    for new organizational strategies or practices.

    4.4 Performance analysis of case retrieval

    To test the feasibility of the proposed CBR method, we

    collected 261 ECRs from previous past development 

    ..

    X8

    L2.0

    2003

    Europe

    FrontSuspension

    F301

    Screwing

    Pilot production

    ..

    X67

    F2.0

    2000

    Asia

    FrontSuspension

    F102

    Fitting

    Pilot production

    -

    0.9678

    0.9098

    0.3358

    0.5118

    0.214---Solution

    0.0352.3025Pilot productionYesProcess

    0.0942.3025LuxuryNoProduct

    0.4472.9957Front SuspensionYesComponent

    0.211.2039AssemblyNoProblem

    Maximum

    Similarity

    Value

    Closest Common

    Parent

    Identical

    wi,

    weight of

    ith

    ontology

    Similarity between conceptsOntology

    0864.1

    035.0094.0447.021.0

    3025.29678.0035.03025.29098.0094.09957.23358.0447.02039.15118.021.0

    =

    +++

    xx +xx+xx +x x=Similarity

    i f 

    Product

    Component

    Problem

    Process

    Case 1 Case 2

    Fig. 8   Similarity calculation between cases

    Inf Syst Front (2006) 8:375 – 394 385

  • 8/20/2019 Capturing and Reusing Knowledge

    12/20

     projects of the automobile company. Since the engineering

    change cases collected were ECRs, this analysis could not 

    use a solution ontology. Through discussion and interaction

    with the engineers at the company, four domain ontologies

    were built using Protégé, an ontology editor used to build

    Semantic Web content. We simulated engineering changecase retrieval using the following methods: concept-based

    retrieval (i.e., process-based, problem-based, and compo-

    nent-based retrievals), ontology-based retrieval, and key-

    word-based retrieval.

    &   Concept-based retrieval : This retrieval mechanism is

    similar to sorting through categorized engineering

    changes using concepts, such as engineering changes

    made to the same component or ones caused by the

    same problem.

    &   Process-based retrieval : This mechanism searches for 

     past engineering change cases that took place during the

    same product development process as the target case,

    and retrieves the recent cases.

    &   Problem-based retrieval : This mechanism searches for 

     past engineering change cases that took place because

    of the same problem as the target case, and retrieves the

    recent cases.

    &   Component-based retrieval : This mechanism searches for 

    cases of past engineering change in the same engineer-

    ing component as the target case. Engineers browse

    component hierarchies to find relevant cases. For 

    example, engineering changes in the form or relocation

    of a heat protector to reduce interference with other 

     parts belong to the same component category.

    &   Ontology-based retrieval : As described in Section   4.3,

    the ontology-based retrieval mechanism determines the

    semantic similarity between complex objects in ontol-

    ogies. This mechanism uses formula (1) to calculate the

    similarity and weight values described in Fig.  6.

    &   Keyword-based retrieval : The basic concept of keyword-

     based retrieval is to find engineering change cases that 

    have keywords similar to the target case. Vector space

    models are very common in information retrieval and

    keyword-based retrieval (Salton & McGill,   1983). They

    represent each object as a vector of features in a   k -

    dimensional space and compute similarity using mea-

    sures such as the cosine or Euclidian distance. We

    extracted keywords from all the cases and eliminated

    keywords with a low term frequency-inverse document 

    Ontology

    Workspace Management

    ActorManagement

    DocumentManagement

    DiscussionManagement

    MilestoneManagement

    CECM

    Process support system

     C  onv  er  s  a t  i   on

     s  u p p or  t  

    Knowledge

    Repository

    Delivery

    Query

    OntologyManagement

    Activity Management

    Role

    Management

    ApprovalRouting

    Reasoned

    SpecifiedStored &Retrieved

    Inde xing

    Case Base

    Fig. 9   Architecture of CECM

    Table 3   Average similarities

     between retrieval methods

    a Significant at alpha = 0.05 b Significant at alpha = 0.10

     N Average similarities Standard deviation (1) (2) (3) (4)

    (1) Process-

     based

    70 2.652381 0.9683   – 

    (2) Problem-

     based

    70 2.761905 1.2176 0.404   – 

    (3) Component-

     based

    70 2.828571 0.9975 0.205 0.668   – 

    (4) Ontology-

     based

    70 3.185714 0.7773 0.000a  0.013a  0.013a   – 

    (5) Keyword-

     based

    70 2.914286 0.9439 0.025a  0.246 0.545 0.073 b

    386 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    13/20

    frequency (TF-IDF) (Salton & McGill,   1983). We

    constructed keyword vectors for each case and calculated

    similarities between cases using the cosine measure.

    The Korean automobile company that we studied uses

    component-based retrieval and simple keyword-based

    search methods. To compare similarities between the case

    retrieval mechanisms, we asked seven expert raters from theautomobile company to rank similarity values and to rate

    relevant cases. We provided ten randomly selected target 

    cases and retrieved the relevant cases using the five methods.

    Table   3   shows the average similarities as rated by the

    experts.

    The ontology-based retrieval method provided the

    greatest average similarity. The average similarities be-

    tween the five methods were tested statistically using

    repeated measures ANOVA. Each pair within the five

    methods was compared. The null hypothesis was that 

    average similarities resulting from the two methods would

     be the same. Table  3 summarizes the comparisons, and thevalues in the cells indicate the significance level. The

    average similarity of the ontology-based method was

    significantly higher than results for the other methods; the

    ontology-based method had a higher average similarity than

    the three concept-based methods (process-, problem-, and

    component-based) at the 0.05 significance level. In addi-

    tion, the ontology-based method had greater average

    similarity than the keyword-based method at the 0.1

    significance level.

    5 Collaborative environment for engineering change

    management (CECM)

    A prototype system called Collaborative environment for 

    Engineering Change Management (CECM) was imple-

    mented, as shown in Fig.   9. CECM consists of three

    subsystems: the engineering change process support 

    system, the knowledge repository, and the ontology

    management. The process support system provides func-

    tions to manage engineering change operations, such as

     processes, activities, conversations, documents, actors,

    discussions, and schedules. The knowledge repository

    enables the representation of engineering change experi-ence and the retrieval of similar knowledge items.

    Annotations on knowledge items and knowledge context 

    are stored in the knowledge repository. The ontology

    management provides functions to edit the existing

    ontology and hierarchies of concepts.

    Fig. 10   Process and activity

    management 

    Inf Syst Front (2006) 8:375 – 394 387

  • 8/20/2019 Capturing and Reusing Knowledge

    14/20

    5.1 Engineering change process and activity execution

    The process and activity management functions in CECM

    attempt to capture both formal and structured knowledge in

    workflows, and informal and unstructured knowledge from

    online ad hoc collaboration. The functions also try to help

    users organize relevant knowledge in engineering changes

    around the engineering change activities because theactivity  is the key element of the   knowledge context   in the

    model.

    After detecting a problem in product development, a

    workspace   is created to handle the engineering change

     process and to organize a temporary team with diverse

     backgrounds. In a workspace, routine activities for 

    approval are predefined, such as initiating an engineering

    change request, evaluating the requested change, and

    issuing engineering change orders, and users can define

    any other activities according to their own needs. An

    activity can be decomposed into sub-activities and can

    have participants, documents, discussions, milestones, and

    conversations.

    Figure 10 shows an example of a workspace concerned

    with   ‘noise reduction on heat protector ’   in the   ‘X8’   car.

    Employees can organize a workspace according to the

    specific requirements of an engineering change and online

    discussion sessions can be created on the workspace that enable the connection of informal knowledge and formal

    documents or outputs.

    5.2 Initiating a draft of an ECR 

    A workspace in CECM generates a workflow for handling

    engineering changes, as shown in Fig.  11, when a user fills

    out the ECR form. Users can associate various types of 

    knowledge with workflow forms, such as documents,

    Fig. 11   Initiating an ECR 

    388 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    15/20

    discussions, and cases retrieved from the repository (see the

    ‘Relating entities’  row of the form in Fig.  11). In this way,

    the workflow is also associated with the relevant knowl-

    edge items from the workspace.

    5.3 Knowledge delivery

    When a user prepares to initiate ECRs or ECOs, similar 

    ECRs, ECOs, and knowledge can be delivered by the

     proactive execution of the CBR mechanism. Figure   12

    shows an example screen in which relevant past cases were

    listed. The user can set the weights of the five ontologies

    for the CBR using two options: weights assigned by the

    user or weights generated by the AHP (default weights).

    When the user selects a relevant case, such as   ‘ Noise

    reduction on climate control systems,’  the context informa-

    tion from the case is presented in the manner shown in

    Fig.   13. The user can see the activities that created the

    knowledge, discussions, and associated decision-making.

    5.4 Evolution of the engineering change document 

    ECRs and ECOs are the final outputs in the workflow-

     bas ed ECM sys tem, sin ce eng ine ering cha nges are

    requested after engineers discuss problems informally. The

    initial ECRs and ECOs in CECM are interim requests and

    evolve gradually through discussions and collaborative

    activities in the workspace. This mechanism of the early

    evolution of the change saves time and costs.

    Figure 14  shows an example of a conversation leading

    to revision of an ECR. If users want to revise the request 

    or the solutions to an engineering change case, they can

    initiate a conversation process. They describe their ideasfor revision and reference entities, especially past 

    engineering change cases, using the conversation form.

    In this way, the evolutionary history of the documents is

    also captured, together with the related discussions and

    decision-making.

    5.5 Collaborative activities

    An activity can be divided into two types: routine and non-

    routine. Routine activities are related to the approval

     processes involved in engineering change management, such

    as initiating an engineering change request, evaluating therequested change, and issuing engineering change orders.

    Users can define any activities used to perform collab-

    orative tasks, such as proposing alternatives and identifying

    the scope of the problem. These activities can be decom-

     posed into sub-activities with participants, documents,

    discussions, milestones, and conversations.

    Fig. 12   Knowledge delivery

    Inf Syst Front (2006) 8:375 – 394 389

  • 8/20/2019 Capturing and Reusing Knowledge

    16/20

    5.6 Closing the workspace

    After the engineers come up with the final solution, it 

    h as to b e app ro ved b y man ag emen t o r the cost  

    management department. The approval of the ECO is

    routed automatically based on the scope of the change.

    Before closing the workspace, the workspace manager 

    can rearrange the workspace by organizing knowledge

    items, removing redundant items and activities, and

    changing the structure of the workspace for later use.

    Knowledge items in the workspace are stored with their 

    contextual information such as related activities, evolu-

    tionary history, related conversations and discussion,actors, and documents, for reuse.

    6 Conclusion

    This paper presented a model of ECM that focuses on

    the integration of collaborative activities and knowledge

    management throughout the lifecycle of engineering

    changes. A prototype ECM system called CECM was

    developed based on the model to demonstrate its

     practical feasibility.

    There are several significant advantages of the

     proposed model. First, the model provides a basis for 

    the integration of informal and unstructured off-line

    collaboration with structured online workflows so that 

    valuable knowledge can be captured effectively in

    context. Second, the collaboration model demonstrated

    how Semantic Web technology can help represent and

    share various types of engineering change-related knowl-

    edge in context. Five types of ontologies were developed

    for the automobile industry, which can be extended or 

    modified for other industries. Third, in order to store,search, and retrieve engineering cases efficiently, the

    CBR technique was used along with the concept-based

    similarity measure so that manual efforts in managing the

    cases could be minimized. A simple experiment showed

    that the CBR-based mechanism provides greater perfor-

    mance in retrieving relevant cases compared with the

    keyword-based search used in many existing ECM

    systems.

    The proposed model and system can be used in other 

    domains where engineering change processes are complex,

    Fig. 13   Knowledge context of 

    the past case

    390 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    17/20

    knowledge-intensive, and costly as well. Such domains

    may include consumer electronics, motorcycle manufac-

    turers, and computer manufacturers.Although many large-scale manufacturers are already

    using their own ECM systems and it may require a huge

    investment to implement CECM, the suggested ECM

    model and CECM present a vision of an advanced ECM

    system for the future that utilizes next generation web and

    intelligent information technologies. We also believe that 

    the huge costs of engineering changes in many industries

     justify such a transition to and investment in the advanced

    model of ECM.

    Acknowledgment   The authors acknowledge the help of the many

    interviewees at the host Korean automobile company in conductingthe case study and research survey.

    Appendix A. Brief introduction to the AHP method

    used to calculate the weights of the ontologies

    We have   O1,   O2, ...,   On   as the criterion for which weight 

    values should be calculated. Using the n  criterion, an (n×n)

    matrix   A   can be defined where element   aij   of   A   is the

    experts’  preference of   Oi  over   O  j  when retrieving relevant 

    engineering change cases. It is also assumed that the

    elements of  A  adhere to the following rules:

    Rule 1. If  aij =α , then  a  ji=1/ α ,   α ≠0.

    Rule 2. If Oi is judged to be of equal relative importance as

    O  j , then aij =1,  a  ji=1; in particular,  a ii=1 for all  i.

    The final objective is to calculate the weight value   wi   for 

    each ontology   Oi. Using the above matrix,   wi   can be

    calculated using the following formula:

    wi ¼  1

     lmax

    Xn

     j ¼1

    aij w j    i ¼  1; 2; :::; nð Þ;

    where  1max  is the maximum eigenvalue of matrix A   (Saaty,

    1980).

    A set of pair-wise comparisons of ontologies wascollected from the experts in the target company using

    questionnaires. Matrix   A   was obtained by integrating the

    experts’  preferences as follows:

    Component Problem Process Product Solution

    Component 1 5.593 6.804 3.476 1.197

    Problem 0.179 1 6.082 2.924 1.71

    Process 0.147 0.164 1 0.281 0.147

    Product 0.288 0.342 3.557 1 0.43

    Solution 0.836 0.585 6.804 2.327 1

    Fig. 14   Evolution of an ECR 

    Inf Syst Front (2006) 8:375 – 394 391

  • 8/20/2019 Capturing and Reusing Knowledge

    18/20

    Consequently, we obtained the normalized weights of 

    the ontologies:

    wc; w pb; w pc; w pd ; w s

    ¼   0:447; 0:21; 0:035; 0:094; 0:214f g

    The terms  wc,  w pb,  w pc,  w pd , and  w s   are the weights of the

    component, problem, process, product, and solution ontol-

    ogies, respectively.

    Appendix B. Definition of the similarity measure

    B.1 Definition of ontologies

    Each ontology  Oi is defined as a tree of concept nodes,  C ki(k =1, 2, ...). (Note that when designating a specific

    ontology, we use the terms  O pd ,  O pb,  O pc,  Oc, and  O s   for 

    the product, problem, process, component, and solution

    ontologies, respectively.)

    B.2 Definition of an engineering change case

    An engineering change case   E   j   is defined as the set of 

    concepts that correspond to the concepts in the ontologies,

    i . e . ,   E  j  ¼   akj   akj    2 O pd  [ O pb [ O pc [ Oc [ O s  and  k    ¼

    1; 2; :::;  number of concepts in the caseg.

    B.3 Definition of the similarity between two concepts

    in a single ontology

    The similarity between two concepts   c pi   and   cqi   in   Oi   is

    defined as (Resnik,  1999)

     sim c pi; cqi

     ¼ log N E  j   cri 2  E  j 

     N U ð Þ  ;

    where   cri   is the closest common parent (CCP) to both   c piand  cqi  and  U   is the entire set of engineering change cases

    in the case base.   N ({ E   j |criZ E   j }) is the number of 

    engineering change cases that belong to concept   cri. The

    CCP represents a node in   O I , which is the parent or 

    ancestor of both of the two given concepts located closest 

    to it. (Note that the numerator in the log term can become

    much smaller in large ontologies with greater depth and

    more concepts. For this reason, it is expected that the

    average value of the similarity is proportional to the size of 

    an ontology.) For example, if we calculate a similarity value

     between the   ‘Screwing’   and   ‘Fitting’   concepts in the

     problem ontology in Fig.   6, the parents of the   ‘Screwing’

    concept are   ‘Fixing,’ ‘Assembly,’   and   ‘Problem,’   and the

     parents of the   ‘Fitting’   concept are   ‘Installation,’ ‘Assem-

     bly,’  and   ‘Problem.’  Therefore, the common parents in this

    case are   ‘Assembly’   and   ‘Problem’   and the CCP is

    ‘Assembly’   because   ‘Assembly’   is closer to   ‘Screwing’

    and   ‘Fitting’   than   ‘Problem.’

    B.4 Definition of the compensation factor for reducing size

    effects (Inverse Concept Frequency)

    The compensation factor  f  i  for  Oi   is defined as

     f  i ¼ log

    Pm

     N Omð Þ

     N Oið Þ  :

    (Note that this measure is similar to the inverse document 

    frequency (IDF) measure widely used in information

    retrieval [Salton & McGill, 1983].)

    B.5 Definition of the similarity between two engineering

    change cases

    The similarity between two cases,  E 1  and  E 2, is defined as

    Sim E 1; E 2ð Þ ¼

    Pi

    wi   f  i   simi   c1i; c2ið Þ

    Pi

    wi;

    where   c1i   and   c2i   are the concepts of   E 1   and   E 2,

    respectively, defined in the   Oi   dimension,   c1Z E 1,   c2iZ

     E 2,  c1iZOi , and  c2iZOi .

    References

    Abecker, A., Bernardi, A., Hinkelmann, K., Kühn, O., & Sintek, M.

    (2000). Context-aware, proactive delivery of task-specific infor-

    mation: The knowmore project.  Information Systems Frontiers, 2

    (3/4), 253 – 276.

    Adler, P. S., & Clark, K. B. (1991). Behind the learning curve: A sketch

    of the learning process.   Management Science, 37 (3), 267 – 281.

    Agostini, A., De Michelis, G., & Grasso, M. A. (1997). Rethinking

    CSCW systems: The architecture of Milano. In   Proceedings of   

    the Fifth European Conference on Computer Supported Cooper-

    ative Work , pp. 33 – 48.

    Ahn H. J., Lee, H. J., Cho, K. H., & Park, S. J. (2005). Utilizing

    knowledge context in virtual collaborative work.   Decision

    Support Systems, 39, 563 – 582.

    Alavi, M., & Leidner, D. (2001). Knowledge management andknowledge management systems: Conceptual foundations and

    research issues.  MIS Quarterly, 25(1), 107 – 136.

    Bergmann, R., & Schaaf, M. (2003). Structural case-based reasoning

    and ontology-based knowledge management: A perfect match?

     Journal of Universal Computer Science, 9(7), 608 – 626.

    Berners-Lee, T., Hendler, J., & Lassila, O. (2001). The semantic web.

    Scientific American; May.

    Broekstra, J., Kampman, A., & Harmelen, F. (2003). Sesame: A

    generic architecture for storing and querying RDF and RDF

    schema. In J. Davies, D. Fensel, & F. Harmelen (Eds.),  Towards

    the semantic web: Ontology-driven knowledge management 

    (pp. 71 – 89), West Sussex: Wiley.

    392 Inf Syst Front (2006) 8:375 – 394

  • 8/20/2019 Capturing and Reusing Knowledge

    19/20

    Chen, C., & Huang, C. (2001). A multiple criteria evaluation of high-

    tech industries for the science-based industrial park in Taiwan.

     Information & Management, 41, 839 – 851.

    Clark K. B., & Fujimoto, T. (1991).   Product development perfor-

    mance: Strategy, organization, and management in the world 

    auto industry. Boston: Harvard Business School Press.

    Clark, K. B., & Wheelright, S. C. (1993).  Managing new product and 

     process development: Text and cases.  New York: Free Press.

    Conklin, J., & Begeman, M. (1988). gIBIS: A hypertext tool for 

    exploratory policy discussion.   ACM Transactions on Office

     Information Systems, 6 (4), 303 – 331.

    Decker, S., Mitra, P., & Melnik, S. (2000). Framework for the semantic

    web — an RDF tutorial.  IEEE Internet Computing, 4(6), 68 – 73.

    Delteil, A., Faron-Zucker, C., & Dieng, R. (2001). Learning

    ontologies from RDF annotations. In  Proceedings of the Second 

    Workshop on Ontology Learning , p. 38.

    De Michelis, G., & Grasso, M. A. (1994). Situating conversation within

    the language/action perspective: The Milan Conversation Model.

    In Proceedings of the 5th Conference on CSCW , pp. 89 – 100.

    Golebiowska, J., Dieng-Kuntz, R., Corby, O., & Mousseau, D. (2001).

    Building and exploiting ontologies for an automobile project 

    memory. In   Proceedings of K-CAP ’ 01, Victoria, October.

    Grant, R. (1996). Toward a knowledge-based theory of the firm.

    Strategic Management Journal, 17 , 109 – 122.

    Grant, R. (1997). The knowledge-based view of the firm: Implications

    for management practice. Long Range Planning, 30(3), 450 – 454.

    Huang, G. Q., & Mak, K. L. (1998). Computer aids for engineering

    change control.  Journal of Materials Processing Technology, 76 ,

    187 – 191.

    Huang, G. Q., Yee, W. Y., & Mak, K. L. (2001). Development of a

    web-based system for engineering change management.  Robotics

    and Computer Integrated Manufacturing, 17 , 255 – 267.

    Huang, G. Q., Yee, W. Y., & Mak, K. L. (2003). Current practice of 

    engineering change management in Hong Kong manufacturing

    industries.   Journal of Materials Processing Technology, 139,

    481 – 487.

    Jarvenpaa, S. L., & Leidner, D. E. (1999). Communication and trust in

    global virtual teams.   Organization Science, 10(6), 791 – 815.

    Karvounarakis, G., Magkanaraki, A., Alexaki, S., Christophides, V.,

    Plexousakis, D., Scholl, M., et al. (2003). Querying the semantic

    web with RQL.   Computer Networks and ISDN Systems, 42(5),

    617 – 640.

    Kolodner, J. (1993).   Case-based reasoning . San Francisco, CA:

    Morgan Kaufmann Publishers.

    Loch, C. H., & Terwiesch, C. (1996). Accelerating the process of 

    engineering change orders: Capacity and congestion effects.

     Journal of Product Innovation Management, 16 , 145 – 159.

    Lorenzo, M. M. G., & Perez, R. E. B. (1997). A model and its

    different application to case-based reasoning.  Knowledge-Based 

    Systems, 9(7), 465 – 473.

    Maus, H. (2001). Workflow context as a means for intelligent 

    information support. In Akman, V. et al (Eds.),   CONTEXT 

    2001, LNAI, Vol. 2116, pp. 261 – 274.

    May, A., & Carter, C. (2001). A case study of virtual team working inthe European automotive industry.   International Journal of   

     Industrial Ergonomics, 27 , 171 – 186.

    May, A., Carter, C., & Joyner, S. (2000). Virtual team working in the

    European automotive industry: User requirements and a case

    study approach.   Human Factors and Ergonomics in Manufac-

    turing, 10(3), 273 – 289.

    Melnik, S. (2000). Representing UML in RDF. Available at   http:// 

    www-db.stanford.edu/~melnik/rdf/uml/ .

    Melink, S., & Decker, S. (2000). A layered approach to information

    modeling and interoperability on the web. In   Proceedings of   

     ECDL’ 00 Workshop on the Semantic Web, Lisbon, Portugal,

    September.

    Monplaisir, L. (1999). An integrated CSCW architecture for integrated

     product/process design and development.   Robotics and Comput-

    er-Integrated Manufacturing, 15, 145 – 153.

     Nonaka, I., & Konno, N. (1999). The concept of   “Ba”: Building a

    foundation for knowledge creation.   California Management 

     Review, 40(3), 40 – 54.

     Numata, J., & Taura, T. (1996). A case study: A network system

    for knowledge amplification in the product development pro-

    cess.   IEEE Transactions on Engineering Management, 43(4),

    356 – 367.

    Pikosz, P., & Malmqvist, J. (1998). A comparative study of 

    engineering change management in three Swedish engineer-

    ing companies. In   Proceedings of DETC ’ 98, Atlanta, GA,

    USA.

    Ramesh, B., & Tiwana, A. (1999). Supporting collaborative process

    knowledge management in New Product Development Teams.

     Decision Support Systems, 27 , 213 – 235.

    RDF. Resource Description Framework, 1999. Available at   http:// 

    www.w3.org/TR/rec-edf-syntax/ .

    RDFS. RDF Schema, 2000. Available at   http://www.w3.org/TR/rdf-

    schema/ .

    Resnik, P. (1999). Semantic similarity in a taxonomy: An information-

     based measure and its application to problems of ambiguity in

    natural language.   Journal of Artificial Intelligence Research, 11,

    95 – 130.

    Saaty, T. L. (1980).   The analytic hierarchy process. New York:

    McGraw-Hill.

    Salton, G., & McGill, M. (1983).  Introduction to Modern Information

     Retrieval . New York: McGraw-Hill.

    Terwiesch, C., & Loch, C. H. (1999). Managing the process of 

    engineering change orders: The case of the climate control

    system in automobile development.  Journal of Product Innova-

    tion Management, 16 , 160 – 172.

    Wong, S., & Burton, R. M. (2000). Virtual teams: What are their 

    characteristics, and impact on team performance?  Computational 

    and Mathematical Organization Theory, 6 , 339 – 360.

    Wright, I. C. (1997). A review of research into engineering change

    management: Implications for product design.   Design Studies,

    18, 33 – 42.

    Yoo, S. B., & Kim, Y. (2002). Web-based knowledge management for 

    sharing product data in virtual enterprises.   International Journal 

    of Production Economics, 75, 173 – 183.

    Hong Joo Lee   is a post doctoral fellow at the Sloan School of 

    Management, Massachusetts Institute of Technology. He received his

    M.S. and Ph.D. degrees in 1999 and 2006, respectively, from the

    Graduate School of Management at Korea Advanced Institute of 

    Science and Technology (KAIST). His research areas are recommen-

    dation systems, intelligent information systems, and virtual collabora-tion and knowledge management.

    Hyung Jun Ahn is a Senior Lecturer at Waikato Management School

    at the University of Waikato, New Zealand. He received his Ph.D.

    from the Graduate School of Management at Korea Advanced

    Institute of Science and Technology (KAIST) in 2004. His main

    research interests include multi-agent systems, intelligent information

    systems, and e-supply chain management.

    Inf Syst Front (2006) 8:375 – 394 393

    http://www-db.stanford.edu/~melnik/rdf/uml/http://www-db.stanford.edu/~melnik/rdf/uml/http://www.w3.org/TR/rec-edf-syntax/http://www.w3.org/TR/rec-edf-syntax/http://www.w3.org/TR/rdf-schema/http://www.w3.org/TR/rdf-schema/http://www.w3.org/TR/rdf-schema/http://www.w3.org/TR/rdf-schema/http://www.w3.org/TR/rec-edf-syntax/http://www.w3.org/TR/rec-edf-syntax/http://www-db.stanford.edu/~melnik/rdf/uml/http://www-db.stanford.edu/~melnik/rdf/uml/

  • 8/20/2019 Capturing and Reusing Knowledge

    20/20

    Jong Woo Kim   is an associate professor of information systems at 

    the School of Business, Hanyang University, Seoul, Korea. He

    received his M.S. and Ph.D. degrees in 1991 and 1995, respectively,

    from the Department of Management Science, the Department of 

    Industrial Management at Korea Advanced Institute of Science and

    Technology (KAIST), Korea. His current research interests include

    intelligent information systems, e-commerce recommendation sys-

    tems, data mining applications, and business process modeling and

    integration.

    Sung Joo Park  is a Professor of Information Systems at the KAIST

    Graduate School of Management in Seoul, Korea. He holds a B.S.

    degree in Industrial Engineering from the Seoul National University, an

    M.S. in Industrial Engineering from the Korea Advanced Institute of 

    Science, and a Ph.D. in Systems Science from the Michigan State

    University. He has been a senior researcher at the Software Develop-

    ment Center, KIST, and a professor at the KAIST since 1980. His areas

    of research interests include intelligent information systems and the

    application of agent technology to management decision-making.

    394 Inf Syst Front (2006) 8:375 – 394