Semantic hyper-graph-based knowledge representation ...

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/324741724

Semantic hyper-graph-based knowledge representation architecture for

complex product development

Article  in  Computers in Industry · April 2018

DOI: 10.1016/j.compind.2018.04.008

CITATIONS

0

READS

50

4 authors, including:

Some of the authors of this publication are also working on these related projects:

Pharmacy delivery SPD system based on lean method View project

Design and development of hospital medicine monitoring system based on Internet of things technology View project

Zhenyong Wu

Guangxi University

25 PUBLICATIONS   230 CITATIONS   

SEE PROFILE

Wenyan Song

Beihang University (BUAA)

40 PUBLICATIONS   349 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Zhenyong Wu on 25 April 2018.

The user has requested enhancement of the downloaded file.

Contents lists available at ScienceDirect

Computers in Industry

journal homepage: www.elsevier.com/locate/compind

Semantic hyper-graph-based knowledge representation architecture forcomplex product development

Zhenyong Wua, Jihua Liaob, Wenyan Songc,⁎, Hanling Maoa, Zhenfeng Huanga, Xinxin Lia,Hanying Maod

a School of Mechanical Engineering, Guangxi University, Nanning 530004, Chinab LiuGong Machinery Co., Ltd, Liuzhou, Guangxi 545007, Chinac School of Economics and Management, Beihang University, Beijing 100191, Chinad College of Automotive and Transportation, Guangxi University of Science and Technology, Liuzhou, Guangxi, China

A R T I C L E I N F O

Keywords:Product development knowledgeKnowledge representationKnowledge serviceXML topic mapOntology

A B S T R A C T

More and more manufacturing companies are facing challenges in knowledge refining and reusing in stage ofproduct development. To resolve this problem and make the knowledge convenient for acquisition, machine-understandable and human-understandable, this paper proposes a framework of semantic hyper-graph-basedknowledge representation to support the knowledge sharing for the product development. A case study of carheadlamp development is given to validate the feasibility and effectiveness of the proposed method. The resultsbring out that it can help engineers to rapidly and accurately acquire knowledge. In future research, theknowledge recommendation service based on product development process should be considered.

1. Introduction

Product development is an intensive knowledge involved, oftencomplex, fuzzy and iterative process in product lifecycle management[1]. The needs and specifications of the knowledge is further refinedover the period of product development process [2]. An efficientknowledge representation scheme can help the designer to make better-informed decisions with effective computer support tools. In today’sproduct development field, product developers or designers need alarge amount of raw data and information to perform their work.Knowledge representation is very important to convert this raw dataand information into knowledge, which is available to designers [3].There is great pressure on the product developer due to product de-velopment risk and efficiency in managing development resources, notjust for the product but also for the development process. Furthermore,the trend to shorten new product development time to stay competitivehas made the new methods develop fast through the use of concurrentengineering and collaborative product development processes [4],which depends on effective flow and share of knowledge betweenproduct development teams [39]. There is a common view that deci-sions made early in the design process have higher impact on productdevelopment time, cost, and sustainability [5]. In later stages of productdevelopment, it often requires knowledge from the earlier stages [6].

Some researches, which include design rationale systems, product fa-milies, systems engineering, and ontology engineering, pursue to cap-ture information or knowledge from early product development deci-sions, customer requirements and feedback analysis reports, productfunctions and associated physical features. The product developmentknowledge generally exists and stores in management/applicationsystem or engineers’ experiences [7]. Without the experience knowl-edge of domain experts, this kind of experience knowledge cannot beshared among engineers effectively [8].

Product development knowledge exists in technical documents,engineering manuals, design drawings and system databases [9]. It ismostly in structured or semi-structured form and stored in hardmemory or information system that use for knowledge sharing andreuse [10]. In recent years, there have been significant and considerabledevelopments in knowledge representation in product development.Some rule-based methods are not good for users to understand. Thegraph-based methods may lack efficiency for knowledge reasoning andstoring [11]. It is useful to focus on the evolution of product develop-ment research. Then a new knowledge representation method is pro-posed. The knowledge representation method should be machine-un-derstandable, human-understandable and convenient for knowledgeacquisition. Therefore, a unified knowledge representation method isthe premise of product knowledge service. Based on this, we present an

https://doi.org/10.1016/j.compind.2018.04.008Received 9 November 2016; Received in revised form 5 February 2018; Accepted 10 April 2018

⁎ Corresponding author at: Beihang University, No.37, Xueyuan RD, Haidian District, Beijing 100191, China.E-mail addresses: [email protected] (Z. Wu), [email protected] (J. Liao), [email protected] (W. Song), [email protected] (H. Mao), [email protected] (Z. Huang),

[email protected] (X. Li), [email protected] (H. Mao).

Computers in Industry 100 (2018) 43–56

0166-3615/ © 2018 Elsevier B.V. All rights reserved.

T

approach to achieve knowledge representation for product develop-ment. The objective of this paper is therefore to propose a knowledgerepresentation architecture which utilizes semantic hyper-graph tosupport the knowledge sharing throughout the product developmentphase.

The remainder of this paper is structured as follows. We first providean overview of the general framework of the knowledge representationmethod for product development. In Section 2, we give the state-of-artreview. In Section 3, we propose the classification of knowledge andclassification of knowledge representation. In Section 4, we propose theprocess of our approach. We detailly discuss the structure of productdevelopment knowledge-service platform (PDKP), i.e. the definition ofPDKP, the construction of the function and the structure of PDKP, theontology applied in PDKP, and how to construct the relations. Wepropose an example to demonstrate how to integrate the PDKP and alsosome analysis of the approach in Section 5. In Section 6, a comparisonand a discussion are provided. Furthermore, in Section 7, conclusionsand potential work are included.

2. State-of-art review

2.1. Classification of knowledge

Knowledge classification is a necessary step for knowledge re-presentation. In the research field of knowledge management of pro-duct development, knowledge can be classified into the following threedimensions.

The first dimension proposed by Nonaka is that knowledge is clas-sified into explicit knowledge and tacit knowledge. Explicit knowledgeexists in product development documents, problem-solving routines,product function and structure description, computer algorithms,technical and management systems, etc. [12]. Such knowledge consistsof the intellectual platform to design and manufacture the product. Onthe other side, tacit knowledge is embedded in experiences, intuition,unarticulated models or implicit rules [13].

The second dimension classifies knowledge into product knowledgeand process knowledge. Product knowledge includes product require-ments, the mapping relationship between parts and assemblies, pro-duct/part functions, evolution-based design rationale in the productlifecycle. Based on the knowledge management processes and the mainstages of the product lifecycle, the product lifecycle knowledge consistsof customer knowledge, development knowledge, production knowl-edge, delivery knowledge and service knowledge [14].

The third dimension is defined by OECD [15] which clarifies theknowledge into four types: know-what, know-why, know-how andknow-who. This dimension is one of the most important dimensions forthe knowledge-based enterprises and organizations.

2.2. Classification of knowledge representation

Owen and Horváth [12] classify knowledge representation into fivecategories: pictorial, symbolic, linguistic, virtual, and algorithmic.

Table 1 shows the five knowledge representation methods and someexamples respectively.

To support multi-domain knowledge sharing, [40] propose an ob-ject-oriented knowledge representation scheme that allows both up-stream and downstream integration of CAPP, and makes it easilyadaptable for interfacing with other computer integrated manu-facturing modules. [41] present a causal loop model to represent causesand effects of through-life engineering service knowledge on productdesign. There are mainly five tacit knowledge representation methods,i.e. protocol analysis, ethnography, graphic thinking, Kansei en-gineering and image scale. Table 2 shows some representation formswith respect to product development for tacit knowledge in productdevelopment cycle.

The ontology approach is often used in knowledge representation.Ontology is effective in representing the structured knowledge.However, with the development of information technology, especiallythe application of semantic technology and Web service technology,some new methods are provided for knowledge representation.However, the industry requires a more convenient and effective methodfor the product development which involves various types of knowl-edge.

The knowledge representation method should be able to representdifferent types of knowledge resources in the product developmentprocess. The specific knowledge classification depends on the specificrequirements of a company. However, the representation method basedon hyper-graph and ontology can describe the relationships betweenknowledge resources and relationships, which can facilitate knowledgecoding and automation. The XML Topic Map proposed in this paper ismore suitable to the knowledge service environment than othermethods, which can support knowledge using and sharing. Moreover, awell knowledge representation method will support product develop-ment and manufacturing and improve the use of product knowledge innew product development process.

2.3. Analysis of literature

As discussed above, the common knowledge representation methodsinclude that semantic network-based method, neural network-basedmethod, concept maps based-method, ontology based-method, se-mantic Web-based method and topic-maps based method. Table 3shows some previous methods. This article mainly focuses on the pro-duct development in the manufacturing industry. The knowledge re-presentation method requires some new features to adapt to thismanufacturing industry environment. The product developmentknowledge representation model must define and represent this se-mantics for subsequently sharing and using product developmentknowledge.

According to the discussion above, the modeling method of theproduct development knowledge needs considering the semantic andsyntax of the representation constructs. In order to develop such a

Table 1Classification of knowledge representation.

Representation category Example

Pictorial Sketches, Detailed drawings, Chart, PhotographsSymbolic Decision tables, production rules, Flowcharts, FMEA

diagramLinguistic Customer requirements, Design rules, constraints,

Customer feedbackVirtual CAD models, virtual prototypes, multimedia,

AnimationsAlgorithmic Computer algorithms, Constraint solver, Design/

operational procedure

Table 2Classification of tacit knowledge representation.

Representation category Case

Protocol analysis Exploring problem decomposition in conceptualdesign [16], engineering design processes [17]

Ethnography Role of shared artifacts [18], implementinginformation systems [19]

Graphic thinking A sketch-based 3D modeling system [20], Sketchrecognition in interspersed drawings [21]

Kansei engineering Improving consumer affective satisfaction [22], User-centric design [23]

Image scale Parameter-based product form and color design [24],innovative product design [25]

Z. Wu et al. Computers in Industry 100 (2018) 43–56

44

model, a semantic web framework based on XML Topic Map (XTM) isproposed. A set of hyper-graph operations on XTM is developed to re-present the distributed knowledge resources in different product de-velopment process. Following this model, representation architecturesare defined to for product development.

3. Classification of knowledge for product development

Explicit knowledge is well represented by formed information andvirtual reality prototypes. People have been focusing on developingmeans of representing tacit knowledge. Mapping tacit knowledge to aphysical form is a very difficult problem and these mappings are notunique. Since knowledge should be defined as information in context,the representation of knowledge would depend on both the content andthe context of the information. A good product knowledge representa-tion model should have the ability to not only capture knowledgethrough the development process, but also reflect the relevant context.A significant part of research in product development is concerned withknowledge capturing, representation, and reuse.

In this paper, the knowledge of product development is classifiedinto four categories which consists of know-what, know-why, know-how and know-who knowledge. Furthermore, considering the knowl-edge user and product development process, there are three knowledgedimensions in product development stage as shown in Fig. 1. The threedimensions are dimension of product development knowledge, knowl-edge users dimension and product development process dimension.

In the light of dimension of product development knowledge,knowledge is divided into four categories in this stage: know what,know why, know how and know who knowledge. Know-what knowl-edge here includes customer requirements, decision tables, productdevelopment cases and customer feedback. Know-why knowledge in-cludes production rules, design rules, detailed drawings and fishbonediagrams. Know-how knowledge mainly includes CAD model views,FMEA diagram, virtual prototypes, computer algorithms and designprocedures. Know-who knowledge includes multimedia, flow charts,and photographs.

The dimension of product development process mainly consists ofproduct plan, concept design, preliminary design and detail design.

Table 3The previous methods of knowledge representation.

Modeling method Researchers Contribution

Semantic networks-based method Woods [26] and Peters and Shrobe [27] Introduce the mechanism of this method and use the method for knowledge representation in anintelligent environment.

Neural networks-based method Widrow et al. [28] and Kasabov [29] Classify the application scenarios of this method. Implement a neural network-based method inonline knowledge-based learning.

Concept maps based-method Sowa [30] and Novak [31] Analyze knowledge of the real world and map it to a computable form.Ontology based-method Gómez-Pérez and Corcho [32] and Chen

[33]Ontology is integrated with the semantic web. Ontology-based empirical knowledgerepresentation and reasoning.

Semantic Web-based method Berners-Lee et al. [34] and Daconta et al.[35]

Make the web context meaningful and understandable.

Topic-maps based method Cañas et al. [36] and Biezunski et al.[37]

Apply the method in knowledge modeling and sharing environment.

Fig. 1. Classification of product development knowledge.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

45

Each sub-process needs various knowledge such as know-what knowl-edge and know-how knowledge.

The third dimension called knowledge users mainly includes twotypes of user: decision maker and technology developer. The targetusers of product development are persons who make decisions andpersons who conduct technology development. The persons making adecision sometimes include product managers in the company. And thepersons conducting technology development include development en-gineers in product development team.

4. Product development knowledge representation model

4.1. Proposed knowledge representation model for product development

The challenge of constructing such a representation model is designa mechanism in which knowledge can be effectively captured, modeled,represented and shared during the different development stages. Thefollowing section presents a proposed model of product developmentknowledge to be used in product development.

As shown in Fig. 2, a model of product development knowledge isproposed, and it consists of three layers. The knowledge ontology inontology layer contains three objects and the detail is shown also inFig. 2. The mapping is created in XTM layer to link knowledge sourcesand requests. The underlying layer is the knowledge resource layerwhich consists of information and data.

The ontology layer can represent the semantic information in thisframework, which mainly consists of development process ontology,development object ontology, and knowledge object ontology. Theprocess ontology generally includes the concept and property in-formation of product development processes, such as developmentstage, task, activities and some process elements. The developmentobject ontology mainly includes the related concepts and property ofdevelopment object, such as product ID, parts category, design versionsand attribute information. Knowledge object ontology consists of theclassification of knowledge, such as know-what, know-why, know-how,and know-who knowledge.

The instance of product development knowledge ontology is thendesigned according to hyper-graph to form the XTM layer. It representsproduct development knowledge by using an instance of ontology. Italso can connect different knowledge resources which include productobjects, concepts, specifications, processes, and experts.

Resource layer stores knowledge resources refined from productdata and information.

4.2. XML Topic Map and its hyper-graph model

Topic Maps is a standard for the representation and interchange ofknowledge with an emphasis on the information finding method. Atopic map represents information using (1) Topics representing anyconcept from people, countries, and organizations to software modules,

Fig. 2. A knowledge model of product development.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

46

individual files, and events, (2) Associations representing hyper-graphrelationships between topics, and (3) Occurrences representing in-formation resources relevant to a particular topic.

These topics and associations form into a group which is called atopic map. TopicMaps.Org produces the XML syntax for topic mapswhich is a reformulation of topic maps in XML syntax based on XLink.The topic map can be used as a technology for the new semantic web, inwhich data and information are given well-defined meaning, making itpossible for computers and people to use and share more effectively.XML Topic Map (XTM) term definitions are shown in Fig. 3.

A Subject is anything that can be talked about or conceived by ahuman being. A Resource is a Subject that has an identity within thebounds of a computer system. Any other Subject is known as a Non-addressable Subject. There are many types of Non-addressable Subjects.A Class is a Non-addressable Subject. Types of Resource include String,XML Element, XML Attribute, Topic Map, Topic Map Node and TopicCharacteristics. Types of XML Element include< topic>Elementand< association>Element. There are three types of Topic Map Node:

Topic, Association, and Scope. There are three types of TopicCharacteristic: Base Name, Occurrence, and Role.

Generally, the topic map can be considered as a hyper-graph [38]. Ahyper-graph is the generalization of the graph concept in which an edgeis incident to an indeterminate number of vertices. There are threedistinct sets of hyper-graph elements: vertex, edge, and incidence.Every incidence in topic map links one vertex and one edge exactly. Inhyper-graph concept, vertex and edge linked by incidence is called in-cident to each other. In the graph theory, the hyper-graph can also berepresented by a graph as follows: (1) the set of vertices is a union of allthe vertices and hyper edges of the hyper-graph; (2) the set of edgesconsists of all the relations of incidence among vertices and hyper edgesof the hyper-graph.

Semantic network is a network with different types of nodes andlinks that contain semantic information. For knowledge representation,

Fig. 3. Class hierarchy of XML Topic Map.

Fig. 4. Illustration of semantic hyper-graph model.

Fig. 5. Semantic network structure of knowledge.

Table 4Detail information of concepts and semantic links.

ID Class Name

t1 Concept Product lifecyclet2 Concept Product developmentt3 Concept Knowledge modelingt4 Concept Knowledge sharingt5 Concept Knowledge recommendationt6 Concept Product lifecycle knowledge managementl1 Semantic link Part ofl2 Semantic link Referencel3 Semantic link Sequentiall4 Semantic link Subclass

Z. Wu et al. Computers in Industry 100 (2018) 43–56

47

feature, concept, and entity can be represented by the semantic node.Semantic link can represent semantic relationships, such as subclass,instance, and cause-effect, etc. Actually, the semantic network can alsobe regarded as a specific hyper-graph. The mapping relationship be-tween semantic network and hyper-graph can be defined and built. Thevertexes, edges, and incidences in the hyper-graph can be representedby the semantic nodes, semantic links and links between the semanticnodes and the semantic links. In Fig. 4, an illustration of the semantichyper-graph model is shown.

A detailed example of a semantic hyper-graph model is shown inFig. 5. The knowledge contained in Fig. 5 can be refined as follows.Product development is one stage of product lifecycle, and a semanticlink called “part of” exists between the two concepts (i.e. product de-velopment and product lifecycle). Knowledge modeling is the premiseof knowledge sharing. Knowledge recommendation is a subclass ofknowledge sharing. Product lifecycle knowledge management consistsof knowledge modeling, knowledge sharing and knowledge re-commendation.

Table 4 provides detail information of concepts and semantic links.The semantic hyper-graph model of the knowledge representation is

illustrated in Fig. 6. In Fig. 6, t1–t6 denote the concepts, l1–l4 denotesemantic links (l1 denotes Part of, l2 denotes Reference, l3 denotes Se-quential and l4 denotes Subclass), and i1–i9 denote incidences.

As shown in Fig. 5, the weight values represent the importance ofsemantic links. Three weight values can be obtained from the threesemantic links, that is, the link between product development andknowledge modeling, the link between product development and knowledgesharing, and the link between product development and knowledge re-commendation. They are 0.3, 0.5, and 0.2 which reveals that productdevelopment activities get more support from knowledge sharing. Basedon the semantic hyper-graph model, a data structure used for knowl-edge representation can be designed and implemented in knowledgesharing system.

4.3. Product development knowledge ontology

As mentioned in the previous section, product knowledge can bedivided into four kinds, i.e. know what, know why, know how andknow who knowledge. Analyzing the keystones of product devel-opment ontology, the establishment method of knowledge ontology

Fig. 6. A semantic hyper-graph model of knowledge representation.

Fig. 7. Ontology model of development process.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

48

Fig. 8. Development object ontology model.

Fig. 9. Development knowledge object ontology model.

Fig. 10. The four different types of knowledge ontology structure.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

49

is presented. On this basis, taking design processes, design objectsand knowledge objects as the hardcore of ontology, extracting themain concepts and properties from every part, the correspondingontology models can be formed. This structure can help to con-solidate design processes, design objects and knowledge objectstogether, achieve the representation of four kinds of knowledge,and lay the foundation for the following knowledge sharing andreuse.

To give a holistic view of the product development knowledge, the

knowledge ontology framework is divided into three parts: designprocess ontology, design object ontology and knowledge object on-tology. The development process ontology model is shown in Fig. 7.This ontology model consists of some ontology elements as shown inFig. 7.

The development process ontology model includes some im-portant product design flow elements, such as Process_elements,Design_activities, Design_tasks, Participants, Design_guides andGuide_models. The process element consists of Output, Input, Tooland Condition properties. These properties and sub-properties re-present the expression form of development process ontology. Theobject properties of development object ontology mainly includehas_output, has_input, use_tool, has_condition, roots_in, comprises, im-plements, designs and designates. The properties of each major objectin the product development process also have inverse properties.

The development object ontology model is shown in Fig. 8. Thisontology model consists of some important ontology element as shownin Fig. 8. The out parts, standard part and external cooperation partshare a common sub-property which is the supplier. These propertiesand sub-properties represent the expression form of development objectontology.

Specifically, the model of development object ontology containssome product development object properties and elements includingProducts, Components, Output_parts, Standard_parts, External_co_parts. Inthe meantime, in order to define development object ontology, someattached properties are defined and added in, such as Functions, BOM,Structures, Materials, Persons, and so on.

These attributes and sub-attributes represent the form of productdevelopment object. The main object properties in development objectinclude has_key_compenont, has_key_part, has_supplier, has_function,has_BOM, has_structure, has_material, and own_by.

The development knowledge object ontology model is shown inFig. 9. This ontology model consists of some ontology elements,which are development process, development object and knowledgeobject. The knowledge object property consists of four kinds ofknowledge ontology such as know what, know why, know how andknow who. These properties and sub-properties represent the ex-pression form of knowledge object ontology.

The ontology model mainly includes knowledge object in pro-duct development process. Besides product development process anddevelopment object, the ontology model includes four types ofknowledge objects ontology, which are know-what ontology, know-why ontology, know-how ontology and know-who ontology. Theproperty of knowledge object ontology is mainly attributed of has_a.The definition of the four different types of knowledge ontologystructure is shown in Fig. 10.

Fig. 10 shows the structure of know-what, know-who, know-whyand know-how knowledge ontology model.

For the know-what knowledge ontology, these knowledge ontologyobject attributes are Person, Departments, Workgroup, Roles, Projects. Therelationships between objects are_owned_by, belong_to, is_related_with,has_role, and has_participate_in.

For the know-who knowledge ontology, the object attributes mainlycontain Persons, Departments, Workgroup, Roles, Projects. Relationshipsbetween objects are is_owned_by, belong_to, is_related_with, has_role,andhas_participate_in.

For the know-why knowledge ontology, the model structure ismainly composed of Person, Resource, Roles,and Projects. The relation-ships between objects are is_owned_by, belong_to, is_related_with, be_use-d_rule, use_rule, and has_participate_in, has_resource.

For the know-how knowledge ontology, the structure includesPersons, Departments, Structure, Function, Projects. The relationships be-tween objects are is_owned_by, belong_to, is_related_with, has_function, andhas_structure.

Table 5The structure of car headlamp.

Catalog ID Part Name

0 assembly1 sub-assembly1-1 lamp assembly1-1-1 lamp1-1-2 retaining ring1-1-3 gasket1-1-4 screw A assembly1-1-4-1 screw A1-1-4-2 O ring1-1-4-3 plastic gasket1-1-5 screw B assembly1-1-5-1 screw B1-1-5-2 adjusting blade1-1-5-3 O ring1-1-5-4 plastic gasket1-1-6 hot pressing screw1-1-7 nut1-1-8 locating pin assembly1-1-8-1 locating pin1-1-8-2 locating pin gasket1-2 reflector assembly1-2-1 reflector1-2-2 ball pivot screw1-2-3 dimming nut1-2-4 lighting circlip1-2-5 tapping screw1-2-6 lens hood1-3 lens assembly1-3-1 lens1-3-2 panel1-3-3 decorative ring1-3-4 internal lens A1-3-5 internal lens B1-3-6 tapping screw1-4 hot melt glue1-5 rear cover1-6 ventilation cover2 Turn signal assembly2-1 seal ring A2-2 terminal2-3 seal ring B2-4 turn signal holder2-5 insulating bush2-6 insulated rubber tape2-7 wire2-8 wire2-9 sheath2-10 locking plate2-11 seal ring C2-12 seal ring D2-13 terminal2-14 terminal pressing plate3 sidelights assembly3-1 seal ring E3-2 hot melt glue3-3 seal ring F3-4 sidelights holder3-5 contact chip4 bulb 12V55/60W5 bulb 12V21W6 bulb 12V5W7 clamp

Z. Wu et al. Computers in Industry 100 (2018) 43–56

50

5. Case study

In order to test the proposed knowledge representation model ofproduct development, a case study based on a car headlamp hasbeen conducted. The headlamp is a complicated electromechanicalunit, and it is often difficult for designers to understand the pro-blems and issues incurred during the usage of car headlamp. Thiscase study demonstrates how the proposed model helps to representknowledge required by the developers during the developmentstage.

5.1. Background of case study

In the development process of automotive lamps, engineers needa lot of knowledge; the knowledge model must be designed anddeveloped before knowledge retrieval reuse and sharing to productdevelopers. The knowledge requirement of lamp development in-cludes (1) Dimensional requirement: In the process of lamp design,it is divided into the important size and general size. (2) Materialrequirements: Lamp parts mainly use plastic material, metal mate-rial and rubber material; it is also a basic property for lamp de-velopment. (3) Performance requirement: The performance of thecomponents is divided into mechanical properties, high-tempera-ture resistance and corrosion resistance. (4) Assembling require-ment: Description of the existed assembly relationships betweenparts dimension and tolerance.

The lamp mainly can be divided into the assembly, sub-assembly,turn signal assembly, sidelights assembly, bulbs and clips. The structureof car headlamp is shown in Table 5.

Specifically, in the stage of lamp cover development, knowledgerequirements of lamp cover development mainly consists of productdevelopment principle of lamp cover, the main function of the lampcover, design reliability of lamp cover, processing, structure andappearance of the lamp cover. These kinds of knowledge are mainlyknow-what and know-why knowledge. Moreover, development ex-periences of lamp cover are mainly knowledge-how and know-whoknowledge. Selecting overall design of lamp cover, knowledge isclassified in Table 6:

5.2. Representation model for headlamp development knowledge

The prototype of headlamp development knowledge re-presentation model is developed to support knowledge sharing andreuse. As shown in Fig. 11, the resource layer can provide knowl-edge resources such as a database of management/applicationsystem, engineers’ information, the case of completed projects. InXTM layer, the mapping between topics and relationship of themare created to link knowledge source and request. In ontology layer,it provides a part of know-how knowledge ontology and show someproperties reflective bowl design.

5.3. XTM syntax definition based on hyper-graph for headlampdevelopment knowledge

The XML Topic Maps are designed for the knowledge-intensiveautomobile enterprise. It models the distributed knowledge re-source as XML Topic Maps. In automobile manufacturing en-terprise, the XTM knowledge management system manages knowl-edge including topics, associations and incidences. The source ofheadlamp development knowledge is composed of two parts: onepart describes the enterprise’s own information and knowledge (e.g.product cases, product clients, competitors). The other part of in-formation and knowledge comes from the cases. The knowledgeincludes the users, products, services, rules, research report anddocument, and the related internet information.

The following are some examples of the reflective bowl about XML-

based knowledge representation for car headlamp developmentknowledge base.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

51

5.4. Ontology definition for headlamp development knowledge

From the knowledge ontology definition above section, this ontologymodel consists of the development process, development object andknowledge object. Similarly, the headlamp development knowledge on-tology property consists of four types of knowledge ontology such as knowwhat, know why, know how and know who. These properties and sub-properties represent the expression form of knowledge object ontology.

As shown in Fig. 12, taking the lamp cover as development object, thisdesign activity belongs to the molding product design and process design.The relationship between them is In_Part_of. The process of cooperation isthe product molding product design and mold fabrication process planning.There are also other different kinds of knowledge, such as know-what, know-why, know-how and know-who knowledge.

5.5. Demonstrated system

An application system is developed to demonstrate the approachesproposed in this case study. Fig. 13 is the shortcut of the main interface.After the product developer enters the product development knowledgeservice system, the developer selects the automobile root as shown inFig. 13. Then the developer selects the body class, and then selects theheadlamp class to learn that the headlamp has ten parts, including bulk-head, lampstand, shield, decoration strip, lampstand cover, triangular edge,fixture, lamp cover, inner cover and reflective bowl. Finally, the developerselects the reflective bowl and then selects the concept stage from the drop-down menu on top right corner, and selects the know-what knowledge typefrom the drop-down menu to get the knowledge as shown in the right sideof the diagram in Fig. 13.

Fig. 14 shows the knowledge sharing process in a headlamp design. Thisprototype system interface includes knowledge definition, knowledge re-commendation, knowledge retrieval and knowledge service for the cranedesign.

6. Comparisons and discussion

Knowledge representation method in this paper considers both theknowledge recommendation and knowledge sharing, and it is con-ducive to knowledge representation for product development. It cansolve the problem of inaccurate knowledge requirements of engineers,and well describe the relationships between the knowledge resources.

According to the results of the case study, the product knowledgesystem uses the proposed knowledge presentation method to organizeand share knowledge. The proposed knowledge representation methodconsiders semantic information integrated with hyper-graph. From theknowledge modeling technology perspective, comparisons between theproposed modeling technology and existed modeling technologies arerefined in Table 7.

According to Table 7, there are two advantages of the proposedmodeling technology:

(1) It is a multi-level knowledge modeling structure which includesontology layer, XTM layer and resource layer. Additionally, theontology schema is also designed which includes developmentprocess, knowledge object, development object, essential informa-tion and relationships.

(2) The key techniques of the knowledge representation and sharing forproduct development include framework and process of knowledgecapturing. These techniques are developed to achieve hyper-graphbased knowledge representation and knowledge reuse during pro-duct development.

From the perspective of knowledge representation framework,comparisons between the proposed knowledge representation frame-work and others are conducted in Table 8.

According to Table 8, the features of the proposed framework are asfollows.

Table 6Four types of knowledge.

Knowledge Type Name Property Process

Know-what pressure, temperature curve explicit knowledge overall designpressure curve explicit knowledge overall designfuzzy gradation curve explicit knowledge overall designPVT curve explicit knowledge overall designphysical parameter explicit knowledge overall designmechanical properties explicit knowledge overall designFMEA explicit knowledge overall design

Know-why boundary dimension calculation formula explicit knowledge overall designphysical parameter explicit knowledge overall designlength calculation formula explicit knowledge overall designthickness calculation formula explicit knowledge overall designrigidity calculation formula explicit knowledge overall designcavity amount calculation table explicit knowledge overall designexternal dimension explicit knowledge overall design

Know-how working temperature rules tacit knowledge overall designworking time rule tacit knowledge overall designpressure rule tacit knowledge overall designtransparency rule tacit knowledge overall designdustproof rule tacit knowledge overall designhigh-temperature test tacit knowledge overall designlow-temperature test tacit knowledge overall design

Know-who process parameters setting tacit knowledge overall designselecting of materials tacit knowledge overall designappearance feature tacit knowledge overall designDFMEA tacit knowledge overall designdesign experts tacit knowledge overall designusers explicit knowledge overall designdesign engineers explicit knowledge overall design

Z. Wu et al. Computers in Industry 100 (2018) 43–56

52

• This proposed knowledge representation framework has strength inknowledge reasoning and coding. The representation frameworkbased on hyper-graph and ontology can describe the relationshipbetween knowledge resources and relationships, which is con-venient for knowledge coding and automatic reasoning.

• The framework is machine-understandable and human-under-standable. It is convenient for knowledge capture. A unifiedknowledge representation method is the premise of productknowledge service. Especially XML Topic Map proposed in thispaper can fit in the knowledge service environment better than theprevious methods.

7. Conclusion and future perspective

7.1. Conclusion

This paper proposes a framework to represent and share product de-velopment knowledge during the product development phase. This frame-work first classifies the product development knowledge into four types,know-what knowledge, know-why knowledge, know-how knowledge andknow-who knowledge. Then a knowledge representation model based onXML topic map and ontology is proposed. A case study of knowledge re-presentations in headlamp development shows that the proposed methodcan be used to improve product development process.

The main features of the proposed framework are summarized asfollows:

(1) Theoretically, a semantic hyper-graph-based knowledge re-presentation model is proposed toward product developmentknowledge management, which combines the advantages of thehyper-graph in reasoning and coding of knowledge managementand the merits of the semantic architecture model in bringing se-mantic relationship into the representation model.

(2) In practice, the proposed framework can help design engineers toachieve cross-department knowledge acquisition and automaticknowledge reasoning in complex product development, because ithas strengths in effectively representing different types of knowl-edge and describing the relationships between them. With the helpof the proposed knowledge representation model, the design en-gineers will not have to spend a lot of time and energy to search forknowledge as they did before in the process of complex productdevelopment.

7.2. Future research opportunities

To further validate the proposed knowledge representation approach,more empirical studies of different product development projects is neces-sary to be conducted in future. Potential future research directions related tothis study is provided as follows:

Firstly, the proposed knowledge representation model based on se-mantic hyper-graph can be integrated with engineering semantic webtechnologies to support knowledge reusing and sharing among multi-disciplinary engineers in product development. Secondly, feedback

Fig. 11. Part of knowledge model for headlamp development.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

53

Fig. 12. An illustration of lamp cover ontology.

Fig. 13. Illustration of knowledge service system for product development.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

54

mechanisms can be designed for the semantic hyper-graph-based frame-work to improve knowledge quality according to the knowledge user’sfeedbacks. Thirdly, semantic hyper-graph-based method can be applied inother product development projects to represent the implicit and experi-ential knowledge of expert to gain further external validation.

Acknowledgments

The authors thank the editor and the anonymous reviewers for theirhelpful comments and suggestions in improving the manuscript. Manythanks Dr. ZhangWei for his valuable suggestions and help in improving thehyper-graph. The work described in this paper was supported by the NaturalScience Foundation of China (Grant No. 51705436 and 71501006), theNatural Science Foundation of Guangxi Province (no.2016GXNSFBA380184).

References

[1] J. Wortmann, A. Alblas, Product platform life cycles: a multiple case study, Int. J.Technol. Manag. 48 (2) (2009) 188–201.

[2] G.J. Hsu, Y.H. Lin, Z.Y. Wei, Competition policy for technological innovation in anera of knowledge-based economy, Knowl.-Based Syst. 21 (8) (2008) 826–832.

[3] H.S. Jagdev, K.-D. Thoben, Anatomy of enterprise collaborations, Int. J. Prod. Plan.Control 12 (5) (2010) 437–451.

[4] S. Wan, D. Li, J. Gao, R. Roy, Y. Tong, Process and knowledge management in acollaborative maintenance planning system for high value machine tools, Comput.Ind. 84 (2017) 14–24.

[5] J. Veldman, W. Klingenberg, H. Wortmann, Managing condition-based maintenancetechnology: a multiple case study in the process industry, J. Qual. Maint. Eng. 17(1) (2011) 40–62.

[6] P.R. Dean, Y.L. Tu, D. Xue, A framework for generating product production in-formation for mass customization, Int. J. Adv. Manuf. Technol. 38 (11) (2008)1244–1259.

[7] B. Denkena, M. Shpitalni, P. Kowalski, G. Molcho, Y. Zipori, Knowledge

Table 8Comparison between the proposed knowledge framework and others.

Logicality Modularity Reasoning Understandability Coding

The framework in this paper √ Partially √ Partially √Pictorial × × Partially √ ×Symbolic Partially Partially √ × √Linguistic × × Partially √ ×Virtual Partially Partially Partially × √Algorithmic √ Partially √ × √

Notes: “√” indicates the knowledge representation framework has the feature, while “×” denotes that the framework does not possess the feature. “Partially” denotesthat the framework partially possesses the feature.

Fig. 14. Knowledge sharing in product development process.

Table 7Comparisons from the knowledge modeling technology perspective.

Existing methods In this paper

Solution Contribution Solution Contribution

Modeling basedon ontology

Modeling general-oriented

MultilevelknowledgeModels

Contain ontologylayer, XTM layer andresource layer

Modeling basedon semantic

Modeling usingsemanticassociation

Modeling basedon hyper-graph

Add hyper-graphconcept in the model

Z. Wu et al. Computers in Industry 100 (2018) 43–56

55

management in process planning, CIRP Ann.-Manuf. Technol. 56 (1) (2007)175–180.

[8] S.H. Liao, Expert system methodologies and applications—a decade review from1995 to 2004, Expert Syst. Appl. 28 (1) (2005) 93–103.

[9] M. Sandberg, I. Tyapin, M. Kokkolaras, A. Lundbladh, O. Isaksson, A knowledge-based master model approach exemplified with jet engine structural design,Comput. Ind. 85 (2017) 31–38.

[10] G. Lyu, X. Chu, D. Xue, Product modeling from knowledge: distributed computingand lifecycle perspectives: a literature review, Comput. Ind. 84 (2017) 1–13.

[11] E. Francalanza, J. Borg, C. Constantinescu, A knowledge-based tool for designingcyber physical production systems, Comput. Ind. 84 (2017) 39–58.

[12] R. Owen, I. Horváth, Towards product-related knowledge asset warehousing inenterprises, Proceedings of the 4th International Symposium on Tools and Methodsof Competitive Engineering, TMCE 2002 (2002) 155–170.

[13] Nonaka, The knowledge creating company, Harv. Bus. Rev. 69 (1991) 96–104.[14] Z.Y. Wu, X.G. Ming, Y.L. Wang, L. Wang, Technology solutions for product lifecycle

knowledge management: framework and a case study, Int. J. Prod. Res. 52 (21)(2014) 6314–6334.

[15] D. Foray, B. Lundvall, The knowledge-based economy: from the economics ofknowledge to the learning economy, The Economic Impact of Knowledge, (1998),pp. 115–121.

[16] L.A. Liikkanen, M. Perttula, Exploring problem decomposition in conceptual designamong novice designers, Des. Stud. 30 (1) (2009) 38–59.

[17] A. Dong, The latent semantic approach to studying design team communication,Des. Stud. 26 (5) (2005) 445–461.

[18] J. Paay, L. Sterling, F. Vetere, S. Howard, A. Boettcher, Engineering the social: therole of shared artifacts, Int. J. Hum.-Comput. Stud. 67 (5) (2009) 437–454.

[19] T. Hartmann, M. Fischer, J. Haymaker, Implementing information systems withproject teams using ethnographic–action research, Adv. Eng. Inform. 23 (1) (2009)57–67.

[20] L.B. Kara, K. Shimada, S.D. Marmalefsky, An evaluation of user experience with asketch-based 3D modeling system, Comput. Graph. 31 (4) (2007) 580–597.

[21] T.M. Sezgin, R. Davis, Sketch recognition in interspersed drawings using time-basedgraphical models, Comput. Graph. 32 (5) (2008) 500–510.

[22] L.Y. Zhai, L.P. Khoo, Z.W. Zhong, A rough set based decision support approach toimproving consumer affective satisfaction in product design, Int. J. Ind. Ergon. 39(2) (2009) 295–302.

[23] R. Roy, M. Goatman, K. Khangura, User-centric design and Kansei Engineering,CIRP J. Manuf. Sci. Technol. 1 (3) (2009) 172–178.

[24] H.C. Tsai, S.W. Hsiao, F.K. Hung, An image evaluation approach for parameter-

based product form and color design, Comput.-Aided Des. 38 (2) (2006) 157–171.[25] C.C. Lin, D.B. Luh, A vision-oriented approach for innovative product design, Adv.

Eng. Inform. 23 (2) (2009) 191–200.[26] W. Woods, What’s in a link: foundation for semantic network, in: A. Collins, Daniel

G. Bobrow (Eds.), Representation and Understanding: Structure in CogntiveScience, 1975, pp. 35–82.

[27] S. Peters, H.E. Shrobe, Using semantic networks for knowledge representation in anintelligent environment, Pervasive Computing and Communications. (PerCom2003). Proceedings of the First IEEE International Conference on, IEEE, 2003.

[28] B. Widrow, D.E. Rumelhart, M.A. Lehr, Neural networks: applications in industry,business and science, Commun. ACM 37 (1994) 93–105.

[29] N. Kasabov, Evolving fuzzy neural networks for supervised/unsupervised onlineknowledge-based learning, IEEE Trans. Syst. Man Cybern. Part B (Cybern.) 31 (6)(2001) 902–918.

[30] J.F. Sowa, Knowledge Representation: Logical, Philosophical, and ComputationalFoundations, MIT Press, 2000.

[31] J.D. Novak, Learning, Creating, and Using Knowledge: Concept Maps as FacilitativeTools in Schools and Corporations, Routledge, 2010.

[32] A. Gómez-Pérez, O. Corcho, Ontology languages for the semantic web, IEEE Intell.Syst. 17 (2002) 54–60.

[33] Y.-J. Chen, Development of a method for ontology-based empirical knowledge re-presentation and reasoning, Decis. Support Syst. 50 (1) (2010) 1–20.

[34] T. Berners-Lee, J. Hendler, O. Lassila, The semantic web, Sci. Am. 284 (5) (2001)28–37.

[35] M.C. Daconta, L.J. Obrst, K.T. Smith, The Semantic Web: A Guide to the Future ofXML, Web Services, and Knowledge Management, John Wiley & Sons, 2003.

[36] A.J. Cañas, G. Hill, R. Carff, N. Suri, J. Lott, G. Gómez, et al., CmapTools: AKnowledge Modeling and Sharing Environment, (2004).

[37] M. Biezunski, M. Bryan, S. Newcomb, ISO/IEC 13250: 2000 Topic Maps:Information Technology, (2009).

[38] Y. Dong, M. Li, HyO-XTM: a set of hyper-graph operations on XML Topic Map to-ward knowledge management, Future Gener. Comput. Syst. 20 (1) (2004) 81–100.

[39] L. Solano, P. Rosado, F. Romero, Knowledge representation for product and pro-cesses development planning in collaborative environments, Int. J. Comput. Integr.Manuf. 27 (8) (2014) 787–801.

[40] D.N. Sormaz, B. Khoshnevis, Process planning knowledge representation using anobject-oriented data model, Int. J. Comput. Integr. Manuf. 10 (1–4) (1997) 92–104.

[41] T. Masood, R. Roy, A. Harrison, Y.C. Xu, S. Gregson, C. Reeve, Integrating through-life engineering service knowledge with product design and manufacture, Int. J.Comput. Integr. Manuf. 28 (1) (2015) 59–74.

Z. Wu et al. Computers in Industry 100 (2018) 43–56

56

View publication statsView publication stats