ontology fusion in high-level-architecture-based collaborative engineering environments

2 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 43, NO. 1, JANUARY 2013

Ontology Fusion in High-Level-Architecture-BasedCollaborative Engineering Environments

Hongbo Sun, Member, IEEE, Wenhui Fan, Weiming Shen, Senior Member, IEEE, and Tianyuan Xiao

Abstract—In high-level-architecture (HLA)-based distributedheterogeneous collaborative engineering environments (CEEs),the construction of federation object model files is time consum-ing. This paper presents an ontology fusion approach aimingat establishing a common understanding in such collaborativeenvironments. The proposed approach has three steps: ontologymapping, ontology alignment, and ontology merging. Ontologymapping employs a top-down approach to explore all bridgerelations between two terms from different ontologies based onbridge axioms and deduction rules. Ontology alignment adopts abottom-up approach to discover implicit bridge relations betweentwo terms from different domain ontologies based on equivalentinference. Ontology merging generates a new collaboration on-tology from discovered equivalent bridge relations. It adopts anaxiom-based ontology fusion strategy and takes heavy-weightedontologies into consideration. It can find all the explicit andderived interontology relations. In a typical CEE, the proposedapproach has a great potential to improve the efficiency of prepa-ration for HLA-based collaborative engineering processes, reducethe work load for adaptive adjustment of existing platforms, andenhance the reusability and flexibility of CEEs. A case studyhas been conducted to validate the feasibility of the proposedapproach.

Index Terms—Collaborative engineering environments (CEEs),high-level architecture (HLA), ontology, product development.

I. INTRODUCTION

IN INCREASINGLY saturated markets, innovation andproduct development are essential conditions for the sale

of products. Adopting collaborative engineering makes fulluse of several independent product development systems andenhances their abilities at the same time. However, as a matterof fact, collaborative engineering environments (CEEs) arecomplicated and comprise various computer-aided engineer-ing (CAE) systems for collaborative design, simulation, and

Manuscript received October 20, 2010; revised May 22, 2011; acceptedAugust 14, 2011. Date of publication May 1, 2012; date of current versionDecember 12, 2012. This work was supported in part by the Chinese NationalHigh-Tech Research and Development Program (863 Program) under Grant2009AA110302 and in part by the National Natural Science Foundation ofChina under Grant 60874066. This paper was recommended by AssociateEditor W. Pedrycz.

H. Sun was with the Centre for Computer-Assisted Construction Technolo-gies, National Research Council, London, ON N6G 4X8, Canada. He is nowwith the National CIMS Engineering Research Center, Tsinghua University,Beijing 100084, China (e-mail: [email protected]).

W. Fan and T. Xiao are with the National CIMS Engineering ResearchCenter, Tsinghua University, Beijing 100084, China (e-mail: [email protected]; [email protected]).

W. Shen is with the Centre for Computer-Assisted Construction Technolo-gies, National Research Council, London, ON N6G 4X8, Canada (e-mail:[email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TSMCA.2012.2190138

optimization. It involves processes like computer-aided design(CAD) modeling, simulation, and optimization and requiresdata and information like CAD digital models, CAE analysis,and optimization results [1], [2]. When several independentsystems need to be integrated, common understanding amongthese systems is always a challenge.

High-level architecture (HLA) is a general-purpose archi-tecture for distributed computer simulation systems. Its earlydevelopment was sponsored by the U.S. Defense Modeling andSimulation Office. In 2000, it was adopted by IEEE as an inter-national standard IEEE 1516 [3]. In its definition, federation isa named set of federate applications and a common federationobject model (FOM) that are used as a whole to achieve somespecific objectives.

Since federates exist within a federation in the form of dataabstraction, federated integration keeps well the independencyof its participants. This kind of integration is more suitable forand is widely used in integrations of distributed and looselycoupled simulation systems. The owner of each participant doesnot need to worry about exposing too much private information.The federation only defines the interesting domains for givenobjectives and the rules of interoperations. It is a real looselycoupled integration solution. Within a federation, subsystemscollaborate in an indirect way so that the context of interopera-tion can be taken into consideration.

Nowadays, more and more simulation functions have beenadded into collaborative product development [4]. The designof a product can be deemed as a multistep process in whicha set of design goals and requirements is transformed intoa functional system. Simulation functions help these systemsfulfill their design goals and add to their potential values.

When simulation is added into a collaborative product de-velopment environment, there always exist several subsystemsin the same environment with independent design goals. Thesesubsystems may follow different design or management rulesin their respective engineering fields [5].

In an HLA-based CEE, federation execution description filesdescribe the data and information exchange standard of a givensimulation. They are essential to common understanding amongcollaborative systems. Within these files, the construction ofFOM needs multidisciplinary professional knowledge and tech-nologies [6]. It is always time consuming and expensive.

Fortunately, ontology in knowledge engineering is the se-mantic basis of communication among domain entities. It isapplicable to automatic reasoning, knowledge representation,and reuse [7]. Ontology-based approaches have been used toresolve the problem of heterogeneous data and informationintegration [8], [9]. The target of this research is to explore a

2168-2216/$31.00 © 2012 IEEE

SUN et al.: ONTOLOGY FUSION IN HLA-BASED COLLABORATIVE ENGINEERING ENVIRONMENTS 3

new FOM construction method which takes full advantage ofontology technologies.

From the viewpoint of reasoning ability, ontology can bebriefly classified into two categories: light-weighted ontologyand heavy-weighted ontology. Light-weighted ontology doesnot have the ability of reasoning. It is, in fact, a well-organizedvocabulary. While, heavy-weighted ontology has reasoningabilities, such as first-order predicate. It usually includes axiomor rule definitions for reasoning use. In this paper, a semiauto-matic construction method of FOM files is proposed. Because itbuilds collaboration ontology from exchanging data ontologiesof subsystems and explores new bridge relations, it is deemedas a heavy-weighted ontology fusion algorithm. The rest of thispaper is organized as follows. Section II reviews the relatedliterature and analyzes the requirements of this applicationproblem. Section III discusses the theoretical foundations ofthe proposed method. Section IV describes the algorithmssupporting the proposed method. Three algorithms are intro-duced in this paper: ontology mapping, ontology alignment, andontology merging. Section V depicts a typical CEE involvingthree systems to demonstrate the applicability of the proposedmethod. Section VI provides conclusions and a further researchplan concerning this research topic. A brief complexity analysisis also included in this section.

II. RELATED WORK

Although the target of this research is to develop a new FOMconstruction method in order to take full advantage of ontologytechnologies, this task is far from just applying existing ontol-ogy technologies into CEEs. The semiautomatic constructionof FOMs can be considered as an ontology integration problem.Ontology integration is the consequence of ontology heteroge-neousness (syntax heterogeneousness and nonsyntax heteroge-neousness [11]). Ontology heterogeneousness can be classifiedinto four layers: representation, terminology, conceptualization,and semantics. In the representation layer, different representa-tion forms are used, and the representation differences can beresolved by formalization. In the terminology layer, differentterms are adopted, and the term differences can be resolved byterm mapping. In the conceptualization layer, ontology theoryalways takes effect here. Furthermore, the problems of thesemantics layer are hard to resolve [10]. When it comes to CEE,because collaboration participants adopt the same ontologyconstruction tools and language, there is no difference in repre-sentation at all. However, because of multidisciplinary-coupledresolutions, regional distribution of organizations, and variousparticipants, heterogeneousness on the terminology, conceptu-alization, and semantics layers cannot be ignored. That leads toseveral challenges in applying ontology technologies to CEE.

1) There is no well-established domain ontology to use [11].2) Every subsystem is totally equal in position. There is

no kernel subsystem. The merging order of capabilityontologies should not influence the final merging results.A metastructure should be designed to support separateddomain laws and bridge relations.

3) There are significant differences among knowledge rep-resentation methods among the subsystems according

to a different series of domain laws [7], [12]. Thesedifferences cannot be easily eliminated by means of ex-isting ontology technologies. Thus, bridge relations (therelations between related concepts from different repre-sentation systems) need to be preset by domain experts.Therefore, the light-weighted ontology approach is notapplicable here.

All the factors mentioned previously bring difficulties whenapplying existing methods to this problem. Most well-knownontology integration tools are not applicable here, includingPROMPT [12], OntoMerge [13], MAFRA [14], GLUE [15],and OntoMap [16]. Some of them are built on literal-basedsimilarity computing methods (e.g., OntoMerge, PROMPT,ONION, and Anchor-PROMPT), while others are too simpleand weak in their description abilities (e.g., OntoMap). Someare instance-based merging (e.g., GLUE); others only adoptbridge axioms. There are also some methods that only taketerms and structures of light-weighted ontologies into consid-eration [17]. When it comes to ontology construction, althoughformal concept analysis [18] can be successfully applied toontology construction, in the procedure of ontology fusion, itis not convenient to use. If this method is adopted, the formalbackground must be recomputed before ontology merging. Thiscomputation brings too much work to CEE, and the computa-tion of multidisciplinary formal backgrounds is difficult.

The main idea presented in this paper is to take full advantageof formalization to automatically discover implicit and explicitbridge relations among term pairs of different ontologies, thento establish the union of the equivalent term pairs as thecollaborative ontology, and, at the same time, to construct awell-defined metaontology (MO) in order to facilitate interoper-ations among independent subsystems. The process of buildingthe output collaborative ontology from independent domainontologies is called CEE ontology fusion.

The main challenges in ontology fusion include instanceand concept confusion, top concept correspondences, modelinghabit differences, synonyms, and coding formats [11]. In anHLA-based CEE, the instances are created by restrict defini-tions of exchange information format. The instances are notconfused with concepts. The foundation of this method is notliteral semantic distance, so the problem of using similar wordshas no effect here. Since the data types used are defined inMO, the coding format is unique. Because the top concept isthe given product, there is no doubt that the top concept isunique. In the scope of one collaboration project, collaborationontologies are built in the same way by the same group ofpeople, so there are no differences in modeling habits. Lastly,no matter what approach is adopted, synonym problems alwaysneed to be addressed by domain experts.

III. THEORETICAL FOUNDATION

Can we use ontology as common understanding media fordifferent disciplinary systems? Theoretically speaking, we can.Because, on one hand, the objective of CEE at one time isunique, every established model is describing some aspects ofthe same thing. On the other hand, the theory of concept lattice(Galois lattice) [19] provides solid ground for this method.


A. Definitions

Because the algebraic system defined on the concept set ofCEE and the partial-order relations of these concepts have thesame upper bound and lower bound, it can be deemed as aconcept lattice [20]. The common understanding models ofCEE have a common source, a given product model, and anymodel involved is specialized in some aspects. They also sharecommon metadata, binary stream, and any datum collaborationrequired is a given parse of a binary fragment. That is to say,the partial-order relations such as part-of or inherit-from have acommon ancestor, the product (�), and all the products have acommon ancestor, Thing. Also, the minimum original concept(⊥) defined under these partial orders is binary characters, andit is also the public descendants of these concepts. Based onthis, this paper defines related concepts as follows.

Definition 1: CEE Ontology:

O ::= (C,HC , RC , HR,M,RM , A).

CEE ontology O is defined as seventuple. C denotes acollaboration concept set of CEE. HC defines a set of partialorders on concept set C. They are inheritance relations amongthe concepts involved. The concept sets and the inheritancerelations form a directed acyclic graph (DAG) whose sourceis the given model of collaborative product and whose sink isa binary fragment. RC denotes a set of noninherited partial-order relations on concept set C, and these partial-order re-lations correspond to concept attributes. HR defines inheritedrelations on the partial-order relation set RC . M is a seriesof collaborative product MO concepts that give a series ofinheritable instances of RC . RM denotes a set of partial-orderrelations under M ; they describe the relations among elementsin the MO set, and they are also the basis for collaborativeproduct ontology reasoning. A defines a set of axioms amongthe ontology concept set and MO relation set, and they providethe major premises of CEE ontology reasoning.

Definition 2: CEE Ontology Fusion:

fuse ::=SETO⇀O, (∀c, c ∈ O→∃f1, f2, f1⇔f2, f1∈Oi1 , f2∈Oi2 , Oi1 ⊂SETO, Oi2 ⊂SETO : fuse(f1, f2,SETO)=c) .

CEE ontology fusion is a partial-order mapping from anontology set of CEE to one ontology (as Fig. 1 shows). Becausethe fused ontology is for collaboration, only more than oneontology (representing the exchange information needed bya subsystem) employs the same concept, and it is useful forfuture use. To any term c in the resultant ontology, it can find acorresponding term f1 in one ontology Oi1 which can be foundin the prepared ontology set. At the same time, to use the termf1, there must be an equivalent term f2 in another ontologyOi2 of the prepared ontology set. Even if the term is unique indifferent ontologies, it will appear in the fused ontology in adifferent name (with prefix of the original ontology). The onlyresults of ontology fusion can be one ontology or null.

The CEE ontology fusion procedure involves mapping froma set of ontologies {O1, O2, . . . , Om} to one collaborationontology Oc. During this process, expert instructions work as

Fig. 1. CEE ontology fusion.

the mechanism, and MO controls the whole ontology fusionprocess.

The ontology fusion method introduced here includes threesteps: mapping, alignment, and merging.

Definition 3: CEE Ontology Mapping:

map ::= EO ⇀ EO, (∀e ∈ O1 → ∃f ∈ O2 : map(e,O1, O2)

= f ∨ map(e,O1, O2) =⊥) .

It is a partial-order mapping based on vocabulary E of oneontology to vocabulary E of another ontology. The term e ∈ Emay be a concept, a relation, or an instance. The mapping maybe an equivalent relation, inherit relation, or ownership. Thismapping is not transmissible except that it describes equivalentrelations. Thus, in the common sense, this mapping does notinvolve more than two heterogeneous ontologies.

Definition 4: CEE Ontology Alignment:

align ::=EO × EO → [0, 1],

align(e, f) = 1 ⇔ e = f : two concepts equal

align(e, f) = 0 ⇔ e �= f : two concepts not equal.

The CEE ontology alignment function involves an equivalentmapping based on the concept vocabulary E of two differentontologies.

Definition 5: CEE Ontology Merging:

merge ::= SETO ⇀ O, (∀e, e ∈ O → ∃f, f ∈ Oi, Oi

⊂ SETO : merge(f,SETO) = e) .

It represents partial-order mapping from an ontology set(SETO) of CEE to one ontology O. Any term e ∈ E in theoutput ontology, no matter whether it is a concept, a relation, oran instance, must have a corresponding term f in one ontologyof the prepared merging ontology set.

An equivalent graph is a directed graph (may be an uncon-nected one). It represents the equivalent relations of concepts inan ontology. In an ontology, it is not possible for two concepts tobe equal (they should be one concept with alias). An equivalentgraph describes equivalent relations between one concept anda group of concepts or attributes. These equivalent relationscan be classified into two categories: structure equivalent anddescription equivalent relations. The structure equivalent re-lation is composed of a group of one-to-many equivalent re-lations defined on the partial-order relations (inheritance) ofthe concepts or their attributes. There are two subcategoriesof structure equivalent relations: inheritance equivalent relationand division equivalent relation (see the definitions hereinafter).


Fig. 2. Equivalent relation denotation.

The description equivalent relation is composed of a group ofone-to-many equivalent relations defined on the partial-orderrelations (ownership) between one concept and a set of itsattributes. One-to-one equivalent relations are included here,which means that an attribute of the concept can be deemed asits identifier. What is more, the description equivalent relationhas quantifier constraints.

Definition 6: Inheritance Equivalent Relation:

Ci ⇔ Cj1 ∩ Cj2 ∩ · · · ∩ Cjm .

The inheritance equivalent relation describes the relation ofa given subconcept Ci that can be uniquely determined by agroup of ancestor concepts {Cj1 , Cj2, . . . , Cjm}. To a givenconcept Ci, there may be no inheritance equivalent relationconcept set or many equivalent relation concept sets.

Definition 7: Division Equivalent Relation:

Ci ⇔ Cj1 ∪ Cj2 ∪ · · · ∪ Cjm .

The division equivalent relation describes a group of com-plete divisions {Cj1 , Cj2 , . . . , Cjm} of concept Ci. That is tosay, any instance of Ci can find a corresponding instance of{Cj1 , Cj2 , . . . , Cjm}. Specifically, the correspondence is notrequired to be unique here. To a given concept Ci, the divisionequivalent relation may or may not exist.

All of these equivalent relations mentioned previously can bedenoted by an m-in-arc with one degree (m is greater than one),as shown in Fig. 2.

Definition 8: Description Equivalent Relation:

Ci⇔ θ1Rk1·Cj1∩ θ2Rk2

·Cj1 ∩ θ3Rkx·Cjy · · · ∩ θmRkn

·Cjm ,

θo=∀|∃|�|=p| ≤ p| ≥ p, p ∈ N+, o ∈ 1, . . . ,m.

This gives the semantic equivalent relations between a con-cept Ci and a group of constrained attributes {θ1, fk1, Cj1}.The constrained attribute group contains a set of attributesconstrained by description logic θoRkx

. Cjy denotes a givenconcept. Cjy denotes the partial-order relation R from Ci toCjy , and θo represents the constraints. The constraints includequantifier constraints and numeric constraints. The universalquantifier constraint ∀ represents that the value range of relationR is all instances of Cjy , the existential quantifier constraint ∃means that relation R has at least one corresponding value ininstances of concept Cjy , the negative constraint � representsthe fact that there is no correspondence of relation R in Cjy

instances, and numeric constraint (= p| ≤ p| ≥ p) means thatrelation R owns a Cjy instance number of = p, ≤ p, or ≥ p,where p is a nonzero natural number. As a matter of fact,

Fig. 3. Mixed equivalent relation denotation.

description equivalent relations and inheritance equivalent re-lations are often used together. Hence, description equivalentrelations are described as follows:

Ci ⇔ Cj1 ∩ θ2Rk2· Cj1 ∩ Cjy · · · ∩ θmRkn

· Cjm .

In an equivalent graph, the constrained attributes are denotedby an identifier on the m-in-arc, as shown in Fig. 3.

An equivalent and mutual exclusive graph is an enhancedgraph G′ based on equivalent graph G with the exclusive rela-tions added (no longer a DAG). The mutually exclusive relationbetween concepts (Ci, Cj) in CEE ontology is a symmetricalrelation, and any instance of Ci and its subconcepts will notbe the instance of Cj and its subconcepts. The equivalent andmutual exclusive graph denotes these relations by ↔ betweenCi and Cj . One mutual exclusive relation may contain anotherone. In that case, two ancestor concepts of mutual exclusionimply descendant concepts of mutual exclusion. This mutualexclusive relation is described as a trivial mutual exclusiverelation.

A structure graph is a graphical representationof inheritance relations HC in a CEE ontologyO ::= (C,HC , RC , HR,M,RM , A). It is a DAG withonly one source, and it denotes the inheritance relations amongconcepts.

Definition 9: Equivalent Relation Bridge:

C(Oi)k ⇔ C

(Oj)l .

An equivalent relation bridge describes the equivalent rela-tion between one term C

(Oi)k in ontology Oi and another term

C(Oj)l of a different ontology Oj . It presents itself in the form

of a term pair (C(Oi)k , C

(Oj)l ). An equivalent relation bridge is

the main concept relation to be found in CEE ontology mappingand ontology alignment.

Domain equivalent bridge axioms refer to a group of seman-tic equivalent relations (C(Oi)

k , C(Oj)l ) from a concept set C(Oi)

of ontology Oi to a concept set C(Oj) of ontology Oj , and theyare the foundation of inference or reasoning in CEE ontologymapping.

B. Mathematical Properties

This paper uses a1 ∧ a2 ∧ · · · ∧ an to denote the maximumlower bound of set {a1, a2, . . . , an} and a1 ∨ a2 ∨ · · · ∨ an torepresent its minimum upper bound.


Fig. 4. CEE ontology fusion framework.

Theorem 1: Operations ∧ and ∨ on CEE ontology con-cept lattice 〈C,�〉 have properties as follows.

1) Idempotent law: For any e ∈ C, there exist a ∧ a = aand a ∨ a = a.

2) Commutative law: For any a, b ∈ C, there exist a ∧ b =b ∧ a and a ∨ b = b ∨ a.

3) Associative law: For any a, b, c ∈ C, there exist (a ∧ b) ∨c = a ∧ (b ∧ c) and (a ∨ b) ∨ c = a ∨ (b ∨ c).

4) Absorption law: For any a, b ∈ C, there exist a ∧ (a ∨b) = a and a ∨ (a ∧ b) = a.

It can be inferred from the aforementioned laws that theminimum upper bound and the maximum lower bound ofany concept in CEE are themselves the concept. This is oneof the most effective ways to align ontologies. When seekingthe minimum upper bound and the maximum lower bound,the same operation is irrelevant to the order. The maximumlower bound of any concept with its ancestors is itself, and theminimum upper bound of any concept with its descendant isalso itself.

Theorem 2: Suppose that 〈C,�〉 is a concept lattice ofgiven CEE ontology; � is the reverse of relation �. For anya, b, c, d ∈ C, the following exist.

1) Reflexivity: a � a; a � a.2) Antisymmetry: a � b and b � a ⇒ a = b; a � b and

b � a ⇒ a = b.3) Transitivity: a � b and b � c ⇒ a � c; a � b and b �

c ⇒ a � c.4) a ∧ b � a, a ∨ b � a, a ∧ b � b, and a ∨ b � b.5) c � a and a � b ⇒ c � a ∧ b; c � a and c � b ⇒ c �

a ∨ b.6) a � b ⇒ a ∧ b = a ⇒ a ∨ b = b.7) a � b and c � d ⇒ a ∧ c � b ∧ d; a � b and c � d ⇒

a ∨ c � b ∨ d.8) Rank preservation: a � b ⇒ a ∧ c � b ∧ c; a � b ⇒

a ∨ c � b ∨ c.9) Distribution inequality: a ∨ (b ∧ c) � (a ∨ b) ∧ (a ∨

c); a ∧ (b ∨ c) � (a ∧ b) ∨ (a ∧ c).10) Norm inequality: a � c ⇒ a ∨ (b ∧ c) � (a ∨ b) ∧ c.

It can be easily inferred from Theorem 2 that the conceptlattice of CEE consists of two Abelian monoids with the samebasic elements. The theorems previously mentioned are animportant source of basic axioms required by CEE ontologyreasoning.

IV. ONTOLOGY FUSION ALGORITHM

As Fig. 4 shows, the main parts of ontology fusion includethree algorithms: ontology mapping, ontology alignment, andontology merging.

The purpose of ontology mapping is to mark definite equiv-alent relations and mutual exclusive relations of any term pairfrom two different ontologies in a collaborative ontology set.These relations are defined at the semantic level, and they canonly be acquired by axioms (universal axioms, user-definedaxioms, and data-type transformation axioms) or deductionsfrom these axioms. The overlapping relations will be inferredand confirmed by the experts in the next step. Different fromquick ontology matching (ontology mapping is a part of ontol-ogy matching) which is based on literal semantic distance, thispaper introduces an equivalent (mutual exclusive) graph-baseddomain axiom (DA) mapping algorithm.

Because (semi) automatic ontology alignment is a key issueof interoperation among ontologies [21], equivalent relationsamong terms should be discovered before interoperations asmany as possible. The objective of ontology alignment is toinfer all the equivalent relations of term pairs in which twoterms are from different ontologies. Because the ontology maynot be constructed in a sufficiently complete way, this procedureneeds domain expert instructions.

CEE collaboration ontology merging is an important basisfor collaboration among several ontologies. It corresponds tothe procedure of negotiations to form a collaborative FOMamong several simulation object models (SOMs) in an HLAframework. This paper introduces an equivalent structure anda concept equivalent bridge-based CEE ontology merging algo-rithm. The main task is to reorganize the equivalent terms whichare found in the procedure of ontology alignment and ontologymapping and then form the collaboration ontology, federationontology (FON; the same as FOM in HLA).

The first step of ontology fusion is ontology mapping. Thiscreates a bridge equivalent concept pair list and the bridgemutual exclusive concept pair list of a given ontology set, basedon DAs. It can be described by algorithm 1.

Algorithm 1. Ontology_mapping({Om}, DA)Input: {Om} candidate mapping ontology setDA domain axiom setOutput: EC bridge equivalent concept pair listIC bridge mutual exclusive concept pair list1 foreach (Oi, Oj) in {Om} do

2 EC ← {(C(Oi)k , C

(Oj)l )}//find domain equivalent bridge

axiom from DA3 IC ← ∅/∗Extract equivalent (mutual exclusive) graphs∗/4 ˜Gi ← Equivalent(Mutual_Exclusive)_Relation_Travel(Oi)

5 ˜Gj ← Equivalent(Mutual_Exclusive)_Relation_Travel(Oj)/∗Simplify equivalent (mutual exclusive) graphs by deletingtrivial equivalent (mutual exclusive) relations (Thing, data-typeequivalent, trivial mutual exclusive relations, and independentconcept nodes)∗/

6 Gi ← Simplify( ˜Gi)

7 Gj←Simplify( ˜Gj)/∗According to {(C(Oi)k ,C

(Oj)l )}, mark

{(C(Oi)k )} of Gi, and mark {(C(Oj)

l )} in Gj , iteratively deleteunmarked concepts of zero in-degree and their m-out-arc ∗/

8 G′i ← Bridge_simplify(Gi)

9 G′j ← Bridge_simplify(Gj)/∗Inferring bridge equivalent

relations, only equivalent graphs of G′i and G′

j , G′=i and G′=

j ,


are used here, and all the discussions below are all based uponstructure equivalent relations∗/

10 foreach unmarked concept Ci in G′=i do

11 if (∃ one–one bridge equivalent relation between ancestorconcepts of Ci and any concept of G′

j , Cj) then12 EC←EC+(Ci, Cj)//duplicate elements eliminated13 elseif (∃ one–one bridge equivalent relation between

ancestor concepts of Ci and any concept of G′j , Cj , and the

attributes, constraints, partial order relations between conceptand its attributes are also equal, the concepts in constraint pathalso have corresponding equivalent bridge concepts.) then

14 EC ← EC + (Ci, Cj)15 end if16 end/∗Inferring bridge mutual exclusive relations∗/17 foreach (C

(OI)p , C

(OI)q ) in {(C(OI)

p , C(OI)q )I ∈ (i, j)} of

Gi or Gj do

18 foreach equivalent concept of C(OI)q , C

(OI)q′ in EC do

19 IC ← IC ∪ {(C(OI)p × C

(OI)q′ )}⊥//According to the

structure graph GOIof ontology//OI{C(OI)

p |GOI}⊥ is concept

C(OI)p and all of its descendents, the//same as {C(OI)

q′ |GOI}⊥.

20 end21 foreach equivalent concept of C(OI)

p , C(OI)p′ in EC do

22 IC ← IC ∪ {C(O)p′ , C

(OI)q }⊥

23 end24 end25 end26 return EC, IC

This algorithm adopts a knowledge representation and atop-down inferring mechanism, based on equivalent (mutualexclusive) graphs and structure graphs. Its inferring abilityis determined by the completeness of the domain equivalentbridge axioms. Compared with most mapping methods usedto date, the main advantage of this algorithm is that most ofthe description features of heavyweight ontology are taken intoconsideration, and this algorithm can find all the explicit andderived bridge relations.

The second step of ontology fusion is ontology alignment,which is developed to search implied bridge relations. Afterconfirmation by the domain experts, these bridge relations areadded into EC for further use.

Algorithm 2. Ontology_alignment({Om}, EC, IC, DTA)Input: {Om} candidate mapping ontology setEC bridge equivalent concept pair listIC bridge mutual exclusive concept pair listDTA equivalent data type axiom setOutput: EC1 foreach (Oi, Oj) in {Om} do//Extract structure graphs2 Gi ← Travel(Oi)3 Gi ← Travel(Oj)

4 {(SC(Oi), SC(Oj))} ← OCi ×OC

j //Cartesian product ofconcept set in Oi and Oj /∗According to mutual exclusivebridge relations, simplify {(SC(Oi), SC(Oj))} ∗/

5 if (∃{(IC(OI)m , IC

(OI)m )} in IC and IC(OI)

m ≡ SC(OI)k )

then//{IC(OI)m |GOI

}⊥ is concept IC(OI)m and all of its descen-

dants//according to the structure graph GOIof OI , the same

as//{IC(OI)m |GOI

}⊥.

6 {(RIC(Oi), RIC

(Oj))} ← {(SC(Oi), SC(Oj))} −{(IC(OI)

m × IC(OI)m )}⊥

7 end if/∗According to equivalent bridge relations, sim-plify {(RIC

(Oi), RIC(Oj))} ∗/

8 if (∃{(EC(OI)m , EC

(OI)m )} in EC and EC(OI)

m ≡RIC

(OI)k ) then//{EC(OI)

m |GOI}� is concept EC(OI)

m and allits ancestors according//to the structure graph GOI

of OI , the

same as {EC(OI)m |GOI

}⊥9 {(REC

(Oi), REC(Oj))} ←

{(RIC(Oi), RIC

(Oj))} − {(IC(OI)m )

� × (IC(OI)m )⊥}

10 end if/∗Inferring equivalent bridge relations.∗/11 {(RMC(Oi), RMC(Oj))} ← {(REC

(Oi), REC(Oj))}

12 foreach (REC(OI)m , REC

(OI)n ) in {(REC

(Oi),REC

(Oj))} do13 if (data-type construction is different accord-

ing to data-type metaclass definition of metaontology)then//the difference of data-type construction includes data-type unit//number inconsistency and data type inheritable

14 {(RMC(Oi), RMC(Oj))} =

{(RMC(Oi), RMC(Oj))} − (REC(OI)m , REC

(OI)n )

15 end if16 end17 {(RUC

(Oi), RUC(Oj))} ← Confirmed({(RMC(Oi),

RMC(Oj))})//domain expert confirmation18 end19 return EC ← EC ∪ {(RUC

(Oi), RUC(Oj)), i < j}

This algorithm is defined on the structure-graph-basedknowledge representation and an attribute-set-comparison-based bottom-up inferring mechanism. The inferring capabilityrelies on the comparison ability between attribute sets. Theoutput of ontology alignment algorithms is term relations, a setof term pairs in different ontologies. This paper proposes theheuristic-information-based (attributes equal) semiautomaticbottom-up ontology alignment method. Because of heuristicinformation involved, the algorithm has a risk of makingwrong judgments, so it needs a domain expert to confirm thecandidate-equivalent bridge relations. It can reach the activeupper bound of implicit equivalent bridge relations throughsearching, which greatly enhances the ability of CEE andremarkably reduces manpower. However, ontology alignmentdoes not generate new ontologies; it only establishes a mappingset to support interoperations among ontologies, so ontologymerging is used here to generate new ontology, based on theexisting ones.

In a typical CEE system, there are always two or moresubsystems. As mentioned previously, the concept lattice ofCEE consists of two Abelian monoids. Thus, the final collab-oration ontology can be constructed by one–one integration inturns.


Algorithm 3. Ontology_merging({Om}, EC)Input: {Om} candidate mapping ontology setEC bridge equivalent concept pair listOutput: FON1 foreach Oi in {Om} do//simplify structure graph of Oi

according to bridge equivalent//concept pair list EC2 Gi → Equivalent_travel(EC,Oi)3 end4 FON≤1 → G1 + Thing (as top concept)/∗Merge Gj ∈

{Gm−11 } and FON≤j−1 in turns∗/

5 foreach Gj in {Gm−11 } do

6 let ECj = {(C(O≤j−1), C(Oj))} in EC

7 let CGj

t = top node of Gj

8 if (∃(CG≤j−1x , C

Gj

t ) in ECj) then

9 Add CGj

t into bridge equivalent concept chain of

CG≤j−1x

10 ECj → ECj − (CG≤j−1x , C

Gj

t )11 else12 Add C

Gj

t into FON≤j−1 as a direct child of Thing13 end if14 foreach node C

Gj

t in Breadth_first_travel (Gj , CGj

t )do

15 if(∃(CG≤j−1x , C

Gj

t ) in ECj) then

16 Add CGj

t into bridge equivalent concept chain of

CG≤j−1x

17 ECj → ECj − (CG≤j−1x , C

Gj

t )18 else19 Add C

Gj

t into FON≤j as a brother node of Gj ,and add all the in-arc of Gj

20 end if21 end22 end23 return FON(FON≤m)

This algorithm is defined on the equivalent-structure-graph-based knowledge representation and an attribute-group-comparison-based merging mechanism. Compared withlightweight ontology merging which is based only on struc-ture and terms, the main advantage of this algorithm is that,when merging ontologies, the heuristic information, such asthe equivalent structure graph and semantic equivalence of theattribute, is also taken into consideration. The efficiency andaccuracy of ontology merging have been greatly improved.

V. CASE STUDY

A typical CEE unit can be described as having three mainparts (Fig. 5): scene creator, localization server, and decisionsupport system. The scene creator generates scenes for col-laboration jobs. The localization server keeps tracking im-portant entities, and the decision support system makes thefinal decision or suggestions about a future collaboration step.The collaboration information among these three independentsystems lies in the scene provided by the scene creator, local-ization information supported by the localization server, and

Fig. 5. Typical CEE unit.

instructions from the decision support system. In this paper,the localization server is supposed to be a real-time trackingsystem, so it does not need any instructions.

Before an HLA-based collaborative job started, the exchangedata formats of each subsystem should be claimed first as SOMfiles (Fig. 6). Then, domain experts sit together to negotiateabout the general collaborative data formats. This process istime consuming and expensive. Then, a general collaborativedata format will be defined, and a FOM file (looks similarto SOM files) will come into being. After these have beenfinished, every subsystem changes its interfaces according tothe new data exchange format. Finally, the collaborative job canbe performed. When the collaborative job changes, this process(negotiation and changing interfaces) repeats again.

When the proposed method is adopted, the workload fornegotiation and interface changing is significantly reduced,although some preparations still need to be done. First, a MO isbuilt to provide templates of capability description ontologies,data-type definitions, and axioms. Then, capability descriptionontologies of these systems are constructed separately accord-ing to the ability and requirements of the related engineeringdomains.

MO is the ontology which stores the fundamental knowledgeof inferences. It contains object templates, basic transformationdata types, and necessary general axioms. This approach iscompatible with the definition of HLA object model templateobject classes and interaction classes.

The scene creator adopted here is a building informationmanagement system. All the files which can be seen in anindustry foundation class (IFC) viewer are deemed as an IFCbuilding. Every element of IFC building has a unique universalidentifier (UUID). IFC building has five subcategories: buildingelement proxy, IFC building elements, single-storey building,multistorey building, and space. Building element proxy is themonitor of building elements, such as hydrometer and thermo-stat. Building element is a part of the building. Single-storeybuilding is a building that only owns one storey. Multistoreybuilding is a building that has at least two floors. Space issomewhere which is isolated. IFC building elements includebasic wall, furnishing elements, opening, roof, and stair. Basicwall with openings is called wall. Opening includes window,door, and door entry. Furnishing elements include fixed furni-ture and movable furniture. The movable furniture is the targetof localization system.


Fig. 6. SOM file of localization system.

The decision support system adopted here is a facilitymaintenance management system. The concepts involved hereinclude point, building description, building element, and build-ing. Point exists in some space; when it has special meaning,it becomes physical location. Building description includesattributes of buildings, such as location and map. Locationcan be specialized into building location, level location, andphysical location. In this system, assets are deemed as buildingelements. Among these assets, assets with locations are lookedas trackable assets. Building is a closed space in the manage-ment scope, which can be specialized into frame, level, or spaceand room. The main objective of this system is asset trackingand maintenance.

The localization system adopted here is called AeroScout.The main concepts used in this system are tag, map, coordinate,asset, corrective action, and building. Tag can be divided into ac-tive tag and inactive tag according to the tracking state. Map canbe specialized into building location, level location, area location,and zone location or physical location. Physical location is also asubclass of coordinate. Asset can be classified into arrangedasset and live asset. Arranged asset belongs to some department.Live asset is a movable asset. Among live assets, trackableassets are binding with active tags. Corrective actions includeevent and internal event, such as invoking and moving. Corre-sponding to the map, building can also be specialized into level,area, and zone. The main targets of tracking are active assets.

The final preparation task is to abstract graphs from theseontologies. Because these systems are too complicated for

a demonstration, only part of the concepts and relations aredescribed in this paper. The bridge relation graphs are shownin Figs. 7–9, and the structure graphs are shown in Figs. 10–12.

In the scene creator equivalent (mutual exclusive) graphdemonstration, there are 12 equivalent relations and three mu-tual exclusive relations.

IFC Building ⇐⇒ product ∩ (= 1 has.UUID) ∩(∃UsingToolkitOf.IFCViewer)

SingleStorey Building ⇐⇒ IFC Building ∩ (=1 has.Storey)

MultiStorey Building ⇐⇒ IFC Building ∩ (> 1 has.Storey)Thermostat ⇐⇒ Building ElementProxy ∩ (=

1MonitorTemperatureOf.IFC Building Element)Hygrometer ⇐⇒ Building ElementProxy ∩ (=

1MonitorHumidityOf.IFC Building Element)Wall ⇐⇒ Basic Wall ∩ (> 1 has.Opening)Space ⇐⇒ Storey ∩ (≥ 4 has.Basic Wall)Room ⇐⇒ Space ∩ (≥ 1 has.Wall)StairWell ⇐⇒ Space ∩ (≥ 1 has.Stair)Fixed Furniture ⇐⇒ Furnishing Element ∩ (=

1FixedAt.Space)Movable Furniture ⇐⇒ Furnishing Element ∩ (=

1LocatedAt.Room)Covering ⇐⇒ Fixed Furniture ∩ (= 1FixedAt.Roof)Single Storey Building ↔ Multi-Storey BuildingFixed Furniture ↔ Movable FurnitureBuilding Element Proxy ↔ IFC Building Element


Fig. 7. Scene creator equivalent (mutual exclusive) graph demonstration.

Fig. 8. Decision support system equivalent (mutual exclusive) graphdemonstration.

Similarly, in the decision support system equivalent (mutualexclusive) graph demonstration, there are seven equivalent rela-tions and one mutual exclusive relation. In the localization sys-tem equivalent (mutual exclusive) graph demonstration, thereare ten equivalent relations and two mutual exclusive relations.

Let S denote the scene creator, L represent the localizationsystem, and D stand for the decision support system. The bridgeequivalent relations given by domain experts are as follows, andno new mutual exclusive bridge relation is inferred

D.Building=S.IFCBuilding D.Level=S.StoreyD.Asset=S.FinishingElement D.LogicalLocation=S.RoomD.Map=L.Map D.Location=L.LocationD.Level=L.Level D.Point=L.PointD.Asset=L.Asset D.Room=L.Zone.

By the ontology mapping algorithm mentioned previously,new bridge equivalent relations are inferred as follows:

D.TrackableAsset = S.MovableFurnishD.PhysicalLocation =L.PhysicalLocationD.BuildingLocation =L.BuildingLocation

D.LevelLocation =L.LevelLocationD.LogicalLocation =L.ZoneLocationD.TrackableAsset =L.LiveAsset.


Fig. 9. Localization system equivalent (mutual exclusive) graph demonstration.

Fig. 10. Scene creator structure graph demonstration.

The candidate-equivalent bridge relations inferred by theontology alignment algorithm are all denied by the domainexperts, and no candidate mutual exclusive bridge relations aregenerated from this algorithm.

After ontology merging, the collaboration ontology isreached, as shown in Fig. 13.

By means of the ontology fusion algorithms described pre-viously, the collaboration ontology is easily established. What

the domain experts need to do is to give well-known equivalent(mutual exclusive) bridge relations and to judge the inferredcandidate relations, which significantly reduces the workloadof domain experts and improves the efficiency of collaborationpreparations. At the same time, because ontologies have ex-pandability and equivalent (mutual exclusive) bridge relationsare also stored in MO for future use, the ability of resource reuseis also remarkably enhanced in the CEEs.


Fig. 11. Decision support system structure graph demonstration.

Fig. 12. Localization system structure graph demonstration.

VI. CONCLUSION

In HLA-based CEEs, it is difficult to establish a collabora-tive system, but it is even more difficult to adaptively adjustinterface codes of existing systems and to negotiate amongmultidisciplinary domains [22]. This paper has proposed asemiautomatic ontology fusion method to establish a collabo-rative ontology as media in HLA-based CEEs, which is con-structed by several independent subsystems. These subsystemsmay be working in different domains. The main part of thismethod includes three algorithms: ontology mapping, ontologyalignment, and ontology merging.

Based on complexity analysis, this approach may not be thebest choice. Because this approach is based on bridge axiomsand equivalent relations, it is not easy to compare with other

Fig. 13. Output collaboration ontology demonstration.

ontology integration methods. For the same reason, OntologyAlignment Evaluation Initiative is not applicable here. Here is abrief complexity comparison. Let n denote the concept numberin the ontology and l represent the length of DA set. Whenintegrating two ontologies, the complexities of existing algo-rithms are as follows: naive ontology mapping O(n2 · log2 n),PROMPT, Anchor-PROMPT O(n · log2 n), GLUE O(n2), andquick ontology mapping O(n · log n) [21]. While, the com-plexity of our ontology mapping algorithm is O(n · l), and thecomplexity of our ontology alignment algorithm is O(n2).

Despite its complexity, it enjoys several remarkable advan-tages which are more suitable for HLA-based multidisciplinarycollaborations.

1) This method has firm theoretical foundations and startsfrom strict definitions.

2) It reduces the workload of domain experts to preparecollaborations among independent engineering domainsby automatic inference and deduction. For the same rea-son, it improves the efficiency in the preparation of futurecollaborations.

3) Different from most other ontology integration methodsusing literature distance, this method employs axioms,bridge axioms, equality rules, and attribute set equalityconditions as the basis for reasoning.

4) The proposed algorithms can find all the explicit andderived bridge relations.

5) Since ontology is used in this method, the reuse ofresources and the expandability of existing systems aregreatly enhanced.

6) The construction method of collaborative ontology alsohelps to accumulate knowledge and, furthermore, to buildrelatively complete models of the same objective fromdifferent aspects.

The applicability of the proposed method has been demon-strated through a case study. More efforts are still requiredin order to improve the proposed method for real industrialapplications.


REFERENCES

[1] W. Fan, W. Wang, and T. Xiao, “Multidisciplinary collaboration simu-lation optimization platform for complex product design,” in Proc. 2ndICPCA, Birmingham, UK, Jul. 2007, pp. 174–178.

[2] H. Sun, T. Xiao, and S. Tang, “Research on federation-based prag-matic integration framework,” in Proc. World Congr. CSIE, Los Angeles/Anaheim, CA, Apr. 2009, pp. 535–539.

[3] IEEE Standard for Modeling and Simulation (M&S) High Level Archi-tecture (HLA)-Framework and Rules, IEEE Std. 1516-2000.

[4] Y. Xu, T. Xiao, C. Liang, and L. Zhang, “Federated integration of net-worked manufacturing service platforms,” Adv. Eng. Inf., vol. 22, no. 3,pp. 317–327, Jul. 2008.

[5] S. Tang, T. Xiao, and W. Fan, “A collaborative platform for complexproduct design with an extended HLA integration architecture,” Simul.Modelling Pract. Theory, vol. 18, no. 8, pp. 1048–1068, Sep. 2010.

[6] IEEE Standard for Modeling and Simulation (M&S) High Level Archi-tecture (HLA)-Object Model Template (OMT) specification, IEEE Std.1516.2-2000, 2001.

[7] W. Zhou, Z. Liu, and H. Chen, “A survey of the research about both FCAand ontology,” Comput. Sci., vol. 33, no. 2, pp. 8–12, 2006, in Chinese.

[8] J. Yu and Y. Dang, “Review on ontology integration,” Comput. Sci.,vol. 35, no. 7, pp. 9–14, 2008.

[9] F. Fürst and F. Trichet, “Axiom-Based Ontology Matching: A Method andan Experiment,” LINA, Nantes, France, Res. Rep. 05.02, Mar. 2005.

[10] C. Bock, X. Zha, H. Suh, and J. Lee, “Ontological product modeling forcollaborative design,” Adv. Eng. Inf., vol. 24, no. 4, pp. 510–524, 2010.

[11] A. Umar and A. Zordan, “Enterprise ontologies for planning and inte-gration of business: A pragmatic approach,” IEEE Trans. Eng. Manag.,vol. 56, no. 2, pp. 352–371, May 2009.

[12] N. F. Noy and M. A. Musen, “The PROMPT suite: Interactive tools forontology merging and mapping,” Int. J. Human-Comput. Stud., vol. 59,no. 6, pp. 983–1024, Dec. 2003.

[13] D. Dou, D. McDermott, and P. Qi, “Ontology translation on the semanticweb,” J. Data Semantics, vol. 3360, pp. 35–57, 2005.

[14] A. Maedche, B. Motik, N. Silva, and R. Volz, “MAFRA—An ontologyMApping FRAmework in the context of the semantic web,” in Proc.Workshop Ontology Transf. ECAI, Lyon, France, Jul. 2002.

[15] A. Doan, J. Madhavan, P. Domingos, and A. Halevy, “Learning to mapbetween ontologies on the semantic web,” in Proc. 11th Int. Conf. WorldWide Web, Honolulu, HI, 2002, pp. 662–673.

[16] H. Schnurr and J. Angele, “Do not use this gear with a switchinglever! Automotive industry experience with semantic guides,” in Proc.4th Int. Semantic Web Wonference, Galway, Ireland, 2005, vol. 3729,pp. 1029–1040.

[17] Y. Huang, H. Zhou, J. Li, L. Tan, and Z. Li, “Survey of ontology mappingapproaches,” Comput. Eng. Appl., vol. 18, pp. 27–29, 2005, China.

[18] B. Ganter, G. Stumme, and R. Wille, Formal Concept Analysis: Founda-tions and Applications. Berlin, Germany: Springer-Verlag, 2005.

[19] R. Wille, “Concept lattices and conceptual knowledge systems,” Comput.Math. Appl., vol. 23, no. 6–9, pp. 492–515, 1992.

[20] K. Qu, J. Liang, J. Wang, and Z. Shi, “The algebraic properties of conceptlattice,” J. Syst. Sci. Inf., vol. 2, no. 2, pp. 271–277, 2004.

[21] M. Ehrig and S. Staab, “QOM—Quick ontology mapping,” in Proc.ISWC, Hiroshima, Japan, Nov. 2004, pp. 683–697.

[22] B. Li, X. Chai, G. Xiong, W. Zhu, C. Quan, H. Zhang, and X. Wang,“Research and primary practice on virtual prototyping engineering ofcomplex product,” J. Syst. Simul., vol. 14, no. 3, pp. 336–341, 2002.

Hongbo Sun (M’10) was born in Fuping, China, onFebruary 13, 1977. He received the B.S. degree ininformation science from Beijing Institute of Tech-nology, Beijing, China, in 1998 and the M.S. degreein software engineering and the Ph.D. degree incontrol science from Tsinghua University, Beijing,in 2005 and 2011, respectively.

During 1998–2002, he was an Assistant Professorwith Shenyang Institute of Technology, Liaoning,China. Between 2005 and 2006, he was a SoftwareEngineer with the National CIMS ERC, Tsinghua

University. From November 2009 to November 2010, he was an InternationalVisiting Worker with the Centre for Computer-Assisted Construction Technolo-gies, National Research Council, London, ON, Canada. He is currently a Post-doctoral Researcher with the National CIMS ERC, Tsinghua University. Hisresearch interests include system integration, artificial intelligence, algorithms,large-scale simulation, and e-commerce.

Wenhui Fan received the B.Sc. degree in materialscience and engineering from Northwestern Poly-technical University, Xi’an, China, in 1990, theM.Sc. degree in material science and engineeringfrom Harbin Institute of Technology, Harbin, China,in 1995, and the Ph.D. degree in mechanical engi-neering from Zhejiang University, Hangzhou, China,in 1998.

He is currently an Associate Professor of controlscience and engineering with Tsinghua University,Beijing, China. Since 1999, he has been working on

complex engineering system design, simulation and optimization, and theirapplications. He has published two books and more than 60 papers in scientificjournals and international conferences and coedited three books in the relatedareas.

Dr. Fan is a member and the Executive Director of the System SimulationSociety of China and the Chairman of the Professional Committee of thesimulation application of the System Simulation Society of China.

Weiming Shen (M’98–SM’02) received the B.Sc.and M.Sc. degrees in mechanical engineering fromthe Northern (Beijing) Jiaotong University, Beijing,China, in 1983 and 1986, respectively, and the Ph.D.degree in systems control from the University of Tech-nology of Compiègne, Compiègne, France, in 1996.

From 1986 to 1992, he was a Lecturer of me-chanical engineering with the Northern (Beijing)Jiaotong University. He has also held positions withMediReport, Paris, France, and the Department ofMechanical and Manufacturing Engineering, Uni-

versity of Calgary, Calgary, AB, Canada. He is currently a Senior ResearchScientist with the Centre for Computer-Assisted Construction Technologies,National Research Council, London, ON, Canada. He is an Adjunct Professorof software engineering with The University of Western Ontario, London.Since 1992, he has been working on intelligent software agents and theirapplications to collaborative engineering design and intelligent manufacturing.He has published three books and more than 300 papers in scientific journalsand international conferences and coedited six books and 18 conference pro-ceedings in the related areas.

Dr. Shen is a member of the American Society of Mechanical Engineers, theAssociation for Computing Machinery, the Association for the Advancementof Artificial Intelligence, and the Canadian Society for Civil Engineering. He isa Registered Professional Engineer in the Province of Ontario, Canada.

Tianyuan Xiao was born in Hunan, China, on Febru-ary 26, 1947. He received the B.S. degree fromthe Department of Electrical Engineering, TsinghuaUniversity, Beijing, China, in 1970.

Since then, he has been active in system modelingand simulation and control theory and application.Since 1993, he has been a Full Professor with Ts-inghua University, where he was the Director of theSystem Integration Research Institute, AutomationDepartment, from 1994 to 2007 and has been the Ex-ecutive Director of the National CIMS Engineering

Research Center since 1994. His research fields cover system simulation, virtualmanufacturing, and Computer Integrated Manufacturing System. He authoredmore than 200 articles and technical papers on Chinese prestigious journals andinternational journals and conferences and published more than ten books, suchas “Application on Computer Simulation,” “Continuous System Simulation andDiscrete Event System Simulation,” “Introduction to System Simulation,” and“Virtual Manufacturing Techniques and Systems.”

Prof. Xiao has been elected the standing member of the Chinese System Sim-ulation Community (CSSC) in 1998, a member of the Board of Directors of theSociety for Modeling and Simulation International in 2004–2006, and the Vice-President of CSSC in 2006. He was a recipient of the National Scientific andTechnology Achievement Award (the second level) for the project “Informationintegration technology and virtual product development (VPD) technology tospeed innovation of textile machinery” (2000), the Chinese CIMS IndustrialLEAD Award(1996) and the Scientific and Technology Achievement Award ofthe National Education Committee (the first level) for the project “CIMS engi-neering of Jingwei Textile Machinery Plant” (1997), the Scientific and Technol-ogy Achievement Award of the National Education Committee (the third level)for the project “The Integrated Manufacturing System Simulator” (1995), andthe National Scientific and Technology Achievement Award (the third level) forthe project “Digital Training Simulator for Operating Ship (1986).”

ontology fusion in high-level-architecture-based collaborative engineering environments

Documents