Advances in Systems Science and Applications (2012) Vol.12 No.1 54-66 The Application of Semantic Web Technology in Manufacturing Grid H.J.Zhang1,2 and Y.F.Hu1,2 1The School of Mechanical and Electronic Engineering, Wuhan University of Technology, Wuhan 430070,China 2Hubei Digital Manufacturing Key Laboratory, Wuhan University of Technology, Wuhan 430070, China Abstract Concerning the knowledge-intensive environment, the paper introduces Semantic Web(SW) technology to Manufacturing Grid (MGrid). Resources and services in MGrid are described in the well-defined meaning, so that computers can un- derstand resource and service information to interoperate seamlessly. In the paper, a semantic-aware MGrid architecture (SAMGA) based on WS-Resource Framework (WSRF) is first presented, in which the stateless Web Services are semantically described using OWL-S and the stateful resource descriptions are semantically enhanced using OWL. Then a semantic matching algorithm is pro- posed for SAMGA. Finally, SAMGA has been applied to the magnetic bearing resource and service sharing platform (MBRSSP). Keywords Manufacturing grid, Semantic Web technology, Semantic matching algorithm, Architecture 1 Introduction Through the network, manufacturing grid (MGrid)[1] makes all kinds of enter- prises and resources geographically distributed connected and form a virtual or- ganization (VO), which is centralized in logic but distributed in physics. In this VO, all resources can be shared and all enterprises can be employed to work collaboratively towards to a common target of dealing with a distributed manu- facturing task or problem. The kernel of MGrid is to share manufacturing resources and use services pro- vided by manufacturing resources pellucidly. However, MGrid resources are far more diverse and complex than those of Grid Computing and there isn’t a u- nified format for describing manufacturing resources and services. There exist lack of relationships and semantics. Moreover, manufacturing service composi- tion is notoriously complex and challenging. These problems seriously hinder the development and application of MGrid. Even though Web Services Description Language (WSDL) was used extensively in Web Service and Grid at present, regrettably, it couldn’t address these problems due to lack of semantic parts. Tangmuarunkit argued that existing resource description and resource selection in the Grid was highly constrained[2]. Due to the complexity of some manufacturing processes, some manufactur- Advances in Systems Science and Applications (2012) Vol.12 No.1 55 ing tasks in MGrid should be decomposed into several subtasks, which cannot be further decomposed and can be executed by a single resource service. Ex- tensive studies have been conducted related to web service composition problem in service-oriented system and distributed system. Existing research efforts for web service composition have been undertaken in two orthogonal directions: 1) manual composition, and 2) automated composition. Manual composition is a time-consuming task, and it is supported by major IT enterprises such as IBM and Microsoft. In this approach, the web service developer selects the outsourced web services that are relevant to their composition intents and programs the in- teraction logic of the component services with a low level programming language such as BPEL[3]. For the automated composition, the information must be un- derstandable by computers, so that they can perform more of the tedious work involved in finding, sharing, and finally combining information in MGrid. The mutual understanding between computers is also important for the optimal com- position of MGrid resources. The appearance of SW[4] brings a new hope of solving the above issues. SW is an evolving extension of the World Wide Web in which the semantics of in- formation and services on the web is defined, making it possible for the web to understand and satisfy the requests of people and machines to use the web content. Many researches have applied SW technologies to grid. All resources and services in grid are adequately described in the well-defined meaning that is machine-processable, which is favor of enabling computers and people to work in cooperation. In the grid system, “semantics” plays an important role in the following aspects: • More effectively discovering and managing dynamic resources in virtual or- ganizations, such as the description and processing of semantic service, resource classification, event notification, origin tracking. The problem of heterogeneity of information in the distributed environment would be resolved at the bottom of the grid architecture; • Linking and cooperating with the information stored and available within a grid automatically, so that it can support the knowledge-intensive applications at the top of the architecture. The rest of the paper is organized as follows. In section 2, some related work is summarized. In section 3, SAMGA is presented. Section 4 introduces a semantic matching algorithm which is especially designed for SAMGA. Section 5 provides the resource and service of magnetic bearing resource and service sharing plat- form based on SAMGA. The conclusion is given in section 6. 56 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid 2 Related Work Early in 2001, a number of researchers were increasingly conscious of the necessi- ty of combination of Semantic Web and Grid. This was firstly captured by David De Roure in the UK e-Science program[5]. Global Grid Forum (GGF) also es- tablished Semantic Grid Research Group (SEM-GRG) to realize the added value of emerging Web technologies and approaches, in particular Semantic Web and Web 2.0, for Grid users and developers. There are some representative research and projects abroad, such as myGrid, CombeChem, GEODISE etc; in China, there is the National Grand Fundamental Research 973 Program: study on basic theory, model and method of Semantic Grid, which would mainly resolve three issues: standard resource organization, semantic interconnection and intelligent aggregation[6]. Above researches are mostly applied in the environment of Computing Grid or Data Grid. There are few researches of Semantic Web technology for manufac- turing. For instance, Yang et al. presented a Semantic Web Services approach for automated integration of manufacturing systems and services[7]. Lemaignan et al. presented a proposal for a manufacturing upper ontology in order to draft a common semantic net in manufacturing domain, and applied the ontology in the automatic cost estimation and semantic-aware multi-agent system for manufac- turing[8]. He et al. proposed an ontology-based manufacturing resource discovery architecture, and gave the discovery algorithm of manufacturing resources based on semantic extending and QoS[9]. Li et al. used ontology to describe resources and services in manufacturing grid, and extend UDDI to support service publish- ing and retrieval[10]. It has been realized that SW have an extremely broad development prospect, but the application in manufacturing just begin, especially in MGrid. Most re- searches have focused on the description of resource and service of MGrid, not on the semantic-enabled MGrid architecture and semantic matching algorithm. Therefore, this paper proposes a new semantic-aware architecture and a semantic matching algorithm for MGrid. 3 Semantic-aware MGrid Architecture Based on concepts and technologies of WSRF and SW, the semantic-aware M- Grid architecture (SAMGA) is designed as shown in Fig.1. For the purpose of the pellucid interoperation, SW technology should be applied upon grid middle- ware. It is important to note that the “semantics” permeates the full vertical extent of architecture and is not just a semantic (or knowledge) layer on top: it is semantics in, on and for the Grid. The function of every layer as follows: Advances in Systems Science and Applications (2012) Vol.12 No.1 57 Fig.1 The semaritic-aware MGrid architecture (1) Fabric Layer: Its basic function is that local resources would be encapsu- lated into global resources which could be shared for the MGrid application layer. At the endpoint of physical resources node, the interfaces of lifetime management, state management and notification would be provided; Web Services standardize various functional operations provided by resources node and offer the standard 58 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid interfaces of access services. Web Services would be described by using service upper ontology OWL-S and manufacturing domain ontologies which are written in standard ontology language OWL. A flexible and high-quality “WS-Resource” would be constructed by combining standard Web Services with dynamic man- ufacturing resources. WS-Resource is the extended grid service according with WSRF specification, hiding the heterogeneity of manufacturing resources and be- ing rich in “semantics”. (2) MGrid core middleware layer: MGrid core middleware is essentially a se- mantic enhanced Web Service container, because the SW standards provide the Web Service container with the capabilities of processing semantic information, such as services composition monitoring, batch job processing and service level agreement (SLA) management. It is the basic running environment of grid ser- vice, which will be deployed in each grid node in advance. It is recommended that MGrid core middleware is developed on Globus Toolkit 4.0.5 core. (3) MGrid public service layer: This is a service set which provides the com- mon operations for MGrid applications, including information service, data ser- vice, task management and semantic service. The semantic service includes three components: • To provide the domain ontology library (knowledge workers who are a little familiar with the knowledge of manufacturing can easily use protg[11] to establish and maintain ontology library) and the general ontology library (e.g., WordNet [12], Cyc[13], Sumo[14]); • To provide the API of ontology libraries (e.g., Jena framework is open source and grown out of work with the HP Labs Semantic Web Programme, which pro- vides an OWL API and SPARQL query engine[15]); • To provide the ontology reasoner (e.g., Racer[16], Pellet[17], FaCT++[18]), OWL-S based matching engine[19] and OWL & OWL-S editor[20]. (4) Manufacturing service layer: On the basis of the MGrid public service lay- er, the intelligent toolkits for the manufacturing applications are developed, for example, CAPP expert system based on artificial intelligence. (5) MGrid application layer: in virtue of the domain dependent program- ming model (e.g. CORBA, COM, JavaBean etc.) and man-machine interaction mechanism, the man-machine interfaces (e.g. visual component Portlet, Servlet, etc.)are provided to MGrid users for MGrid applications. Both OWL-S and the Web Service Modeling Ontology (WSMO) [21] can describe services semantically. OWL-S uses OWL in combination with WSDL. WSMO uses F-Logic to perform inferences with services, and its XML-based format gives external agents access to the service features. The paper adopts OWL-S as MGrid services description language (see Fig. 5). Advances in Systems Science and Applications (2012) Vol.12 No.1 59 4 Procedures of MGrid Resource & Service Publication and Discovery Based on SW 4.1 MGrid Resource & Service Publication Elenius et al. developed an OWL-S plug-in for Protégé as OWL-S editor [20]. The OWL-S editor provides a graphical user interface to create and modify an OWL-S description including all three parts: ServiceProfile, ServiceModel and Service- Grounding. Publishing SW services is more complex than Web Services. Howev- er, it can contribute to automated discovery, negotiation, composition, execution, and monitoring of web services. First of all, RSP develop WSDL documents of Web Services. Then with the help of OWL-S editor, RSP fill in ServiceProfile information and use Service–Model to define service executable process. At last, RSP import the above WSDL documents into OWL-S files. In order to describe MGrid services formally as much as possible, RSP can enter some keywords and browse ontolgoies that include the keywords, as well as synonyms, superclass or subclass of these ontologies, when entering IOPEs. information. Thus RSP select the right ontology names as the IOPEs terms. 4.2 MGrid Resource & Service Discover The procedures of semantic-aware service discovery are shown in Fig.3. RSC just need to enter input and (or) output term(s) via the man-machine interaction interface. In the same way of resource & service publication, after that ontology reasoner performed reasoning and expansion of ontology, RSC select the right ontology names. The selected ontology names and those related terms (e.g., synonyms, superclass or subclass) are sent to MDS4 as keywords. MDS4 connect to all MGIIS and return the Service Profile information to OWL-S based matching engine. Matching engine would extract I/O information from MDS4 and then start to perform the matching algorithm (detailed in section 5) to compare I/O information between returned services and RSC request. Service query interface will show the matching results (including service name and address). Thus RSC connect the selected MGRIS to schedule this service node. 5 Semantic Matching Algorithm The matching algorithm is the key to the implementation of the SW technology in MGrid. The algorithm is the main process of the OWL-S based matching engine in the core middleware layer. The principles of the semantic matching algorithm, which are based on the semantic ontology library-WordNet, are as follows: 60 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid Fig.2 Semaritic-aware MGrid service dicovery model (1) The smaller the semantic distance, the greater the similarity. The semantic distance between one concept and itself is 0, and their similarity is 1. When the semantic distance between concepts is infinite, their similarity is 0. In the hierar- chical structure tree (see Fig.3), the semantic distance refers to the length of the shortest path between two concepts. As shown in Fig.3, the semantic distance of O31 to O36 is denoted by L1 +L2, where L1 refers to the distance of O31 to O11, L2 refers to the distance of O36 to O11 , and O11 is the nearest common concept of O31 and O36. (2)The more the semantic overlap, the greater the similarity. The semantic overlap refers to the level of the same meaning that both concepts involve. As shown in Fig.3, the semantic overlap of O31 to O36 is denoted by L, where L refers to the distance of O11 to O0, and O0 is the top concept in the hierarchical structure tree. Generally speaking, the larger the semantic overlap , the smaller the difference between the meanings of both concepts. For example, the similar- ity between O31 to O32 is larger than that between O11 to O12. (3) The more detailed the classification, the lower the similarity. There are fine and coarse classifications of concepts in the semantic dictionary. Hence, the concept density isn’t a fixed value. There must consider the concept density in the algorithm . (4) The similarity is asymmetric. Generally speaking, the similarity from O1 to O2 is unequal to that from O2 to O1. For example, the similarity from “lathe” to “machine tool” is larger than that from “machine tool” to “lathe”. Because machine tool includes lathe, grinder, shaper etc. Therefore, we introduce the vector curves, such as the broken blue curves with arrow in the Fig.3. According to the above principles, the similarity of resource functionality from concept O1 to concept O2 is defined Sim(O1 → O2) as follows: Advances in Systems Science and Applications (2012) Vol.12 No.1 61 Sim(O1 → O2) = L L+ α(O1, O2) L1 ρ(o) + [1− α(O1, O2)] L2 ρ(o) where α(O1, O2) = { L+L1 2L+L1+L2 , L1 ≤ L2 L+L2 2L+L1+L2 , L1 > L2 is the adjusting parameter of the asymmetry of similarity. The concept density [22] ρ(o) = m−1∑ i=0 nhypi 0.20 descendants0 ,where nhyp refers to the mean number of hyponyms per node, h refers to the height of the subhierarchy, and m refers to the number of senses in the hierarchical struc- ture tree. There is an example of how to cacluate the similarity in the field of machine tool service. The goal of calculating the similarity is to match two resource names. The resource name stands for the resource functionality. The hierarchial structure tree in the WordNet is shown in Fig.4. From the Fig.4, we can obtain the data in the Table 1. Table 1 The similarity calculation O1 → O2 L L1 L2 m desc− α(O1, O2) nhpy ρ(O) Sim(O1 endanto → O2) lathe→ 9 2 0 2 5 0.45000 1.15091 0.43018 0.81138 machine tool grinder→ 9 1 0 2 5 0.47368 1.15091 0.43018 0.89099 machine tool shaper→ 9 1 0 2 5 0.47368 1.15091 0.43018 0.89099 machine tool machine tool 9 0 2 2 5 0.45000 1.15091 0.43018 0.77874→lathe machine tool 9 0 1 2 5 0.47368 1.15091 0.43018 0.88033→grinder machine tool 9 0 1 2 5 0.47368 1.15091 0.43018 0.88033→shaper lathe→ 10 1 1 2 5 0.45000 1.00000 0.66667 0.86957 milling machine lathe→ 9 2 1 2 5 0.47619 1.15091 0.43018 0.72396 grinder In the table 1, it’s easy to see that Sim(lathe → tool) > Sim(lathe → grinder) and Sim(lathe → machine tool)> Sim(machine tool → lathe). The first inequa- tion shows when a lathe is required, the resource node of machine tool is much better for the searching result than that of grinder. The second inequation verifies the asymmetry of similarity. We set a threshold for the similarity in the MGrid 62 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid sysytem. if Sim(O1 → O2) is no less than the value of threshold, we argue that O1 is similar enough to O2. Fig.3 The hierarchical structure tree Fig.4 An example of hierarchical structure tree in wordnet 6 Application A magnetic bearing is a new type of high performance bearing. However, each type of magnetic bearings must be designed and manufactured according to the concrete objects. A large number of resources are needed in the development of magnetic bearing. In order to realize the sharing and collaborative work of all needed resources, according to SAMGA, the magnetic bearing resource and service sharing platform (MBRSSP) is developed. Resource & service providers (RSPs) can publish the own resources through the manufacturing resources pub- lication center, as shown in Fig.4. The publication center provides the semantic template for resource and service publication and automatically generates the se- mantic documents. Resource & service demanders (RSDs) can search the needed resources at the resource optimal-allocation center as shown in Fig.5; RSDs can Advances in Systems Science and Applications (2012) Vol.12 No.1 63 also reserve all kinds of resources for MGrid tasks through the co-reservation sys- tem, as shown in Fig.6. The semantic matching can provide the better searching result for both the optimal-allocation and the co-reservation. Fig.5Manufacturing resource & servicepublication center based on OWL&OWL- S Fig.6 MGrid resource & service optimal-allocation center 64 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid 7 Conclusion This paper presented a semantic-aware MGrid architecture that exploits Semantic Web technologies to solve manufacturing problems. The semantic environment at the bottom of SAMGA can adequately support the knowledge-intensive applica- tion of MGrid on the top layer. With the support of SAMGA, a new information model suitable for the semantic and grid environment is designed, which can re- alize the fuzzy matching. Furthermore, the information model is also called the pull model, which avoids a large number of redundant information. Fig.7 MGrid resource & service co-reservation system Acknowledgements This paper is supported by the National Natural Science Foundation Key Project of China: Digit manufacturing basic theories and key techniques under network environment (NO.50335020), and the Hubei Digital Manufacturing Key Labo- ratory Opening Fund project: Research on resource service search and optimal- selection theories and experiments in manufacturing grid system (No.SZ0621). References [1] Fei Tao, Yefa Hu and Zude Zhou. (2008), “Study on manufacturing grid and its executing platform”, International Journal of Manufacturing Technology and Management, Vol.14, No.1-2, pp.35-51. [2] Hongsuda Tangmuarunkit, Stefan Decker and Carl Kesselman. (2003), “On- tology based resource matching in the grid-the grid meets the semantic web”, Proceedings of the 2nd Inter. Semantic Web Conference, pp.706-721. Advances in Systems Science and Applications (2012) Vol.12 No.1 65 [3] Matjaz B. Juric, Benny Mathew, Poornachandra Sarang. (2004), Business Process Execution Language for Web Services, Birmingham.U.K.: Packt. [4] http://www.semanticweb.org [5] http://www.intsci.ac.cn/ai/sg.html [6] Zhonghhua Yang, Robert Gay, Chunyan Miao, et al. (2005), “Automating integration of manufacturing systems and services: A Semantic Web Services approach”, 31st Annual Conference of IEEE Industrial Electronics Society, pp.2255-2260. [7] Severin Lemaignan, Ali Siadat, Jean Yves Dantan, et al. (2006), “MASON: A proposal for an ontology of manufacturing domain”, IEEE Workshop on Distributed Intelligent Systems: Collective Intelligence and Its Applications, pp.195-200. [8] Yu’an He, Tao Yu, Lilan Liu, et al. (2006), “Research on manufacturing resource discovery based on ontology and QoS in manufacturing grid”, 2006 International Conference on Cyberworlds, pp.209–215. [9] Jing Li, Xiangxu Meng, Shijun Liu, Ran Yuan. (2007), “Ontology-based re- source description in manufacturing grid”, Computer Supported Cooperative Work in Design, pp.646–650. [10] http://protege.stanford.edu [11] http://www.cyc.com [12] http://wordnet.princeton.edu [13] http://www.ontologyportal.org [14] http://jena.sourceforge.net [15] http://www.racer-systems.com [16] http://pellet.owldl.com [17] http://owl.man.ac.uk/factplusplus [18] Naveen Srinivasan, Massimo Paolucci, Katia Svcara. (2004), “Adding OWL- S to UDDI. Implementation and throughput”, Proceedings of the 1st Inter- national Workshop on Semantic Web Services and Web Process Composi- tion.San Diego, California. 66 H.J.Zhang:The Application of Semantic Web Technology in Manufacturing Grid [19] Daniel Elenius, Grit Denker, David Martin, et al. (2005), “The owl-s editor-a development tool for semantic web services”, In Proceedings of the Second European Semantic Web Conference. [20] http://www.wsmo.org [21] Zhizhong Liu, Huaimin Wang, Bing Zhou. (2007), “A two Layered P2P model for semantic service discovery”, Journal of Software, Vol.18, No.8, pp.1922-1932 [22] Eneko Agirre, German Rigau. (1995), “A proposal for word sense disam- biguation using conceptual distance” (Submited), International Conference of Recent Advances in Natural Language Processing. Corresponding author Y. F. Hu can be contacted at: haijun@whut.edu.cn