A Semantic Web Services Based Framework for Enterprise-Wide CollaborativeEngineering Design

Xiaorong Xiang Gregory MadeyDepartment of Computer Science and Engineering

University of Notre DameNotre Dame, IN 46556

xxiang1@nd.edu, gmadey@nd.edu

Abstract

An engineering design process may involve the effortsof many discipline specific design partners. These part-ners may be both cooperative and competitive. In this pa-per, we propose a web services based framework to supportthis process. This framework enables composition of thedistributed work flow for integrating design models from astandard web browser. It also provides a semantic descrip-tion of web services using semantic web technologies. Thegoal for this framework is to facilitate integration of modelsand tools that contain proprietary information in engineer-ing design domain.

1. Introduction

Engineering design processes, such as automobile andaircraft design, involves several partners located at differ-ent locations. These partners may be both cooperativeand competitive. Successful engineering design requireswell-coordinated interactions between individuals or teamsin specialized knowledge domain, information exchange,models, and integration to achieve an optimal goal. How-ever, there is a significant part of design models and toolscontaining proprietary information that cannot be disclosed.Also, these models and tools are normally written in vari-eties of programming languages and run on different plat-forms. With web services technologies, these models andtools can be treated as black boxes and run at their originallocations.

In the integration process, two techniques are involved:exposing part or all of the functionalities as web services(web service management) and developing an applicationthat uses web services (web services composition). In thispaper, we propose a semantic web services based frame-work that supports the collaborative engineering design

process. This framework enables composition of the dis-tributed work flow for integrating design models from astandard web browser. It also allows publishing and man-agement of web services using semantic web technologies.Instead of using the proposed DAML-S and OWL-S stan-dards for the semantic description of web services, we pro-pose an alternative approach that adds one more ontologylayer above the WSDL to describe the properties of web ser-vices. The goal for this framework is to facilitate integrationof models and tools that contain proprietary information inthe engineering design domain.

A view of the collaborative engineering design problemdomain is presented in section 2. Important existing tech-nologies in semantic web services are described in section3. The proposed semantic web services based frameworkand the implementation issues are illustrated in section 4.Conclusions are drawn and future works are described insection 6.

2. Enterprise-Wide Distributed Multilevel De-sign

In Figure 1, the logical view of the system, subsystemand component optimization modules, represent the dis-tributed multilevel design optimization process. The systemconsists of two major disciplines: the engineering disciplineand the business discipline. The distributed enterprise, busi-ness analysis models, and information retrieval services areavailable online for design, planning, and decision makingby end users in the organization. Some of the engineeringanalysis models are enterprise resources and accessible byall authorized users, while other models may be proprietaryto a supplier. Some of the analysis models may be appliedat the system level, while others may be applicable at thesubsystem or component level.

The design process is often a collaborative process thatinvolves multiple specific design teams in the same orga-

Figure 1. Logic View of Distributed MultilevelDesign Optimization

nization or across the organization’s boundary. Integrationof these discipline specific models is a daunting challenge.However, the emerging semantic web services technolo-gies can help the integration and decision making process.In Figure 2, each enterprise wide analysis module and in-formation services module are wrapped as web services.They are “plugged” into the Internet and enterprise Intranet.By using open web services standards, existing legacy andnew applications, are deployed as online interoperable dis-tributed services. An engineering analyst (designer) and anenterprise analyst are able to access any single web servicesmodule. The analysts can compose a collection of individ-ual web services into a composite web service potentiallycomprised of many optimization, analysis, and informa-tion retrieval services. The composite services must be se-quenced so that they execute in the proper order, reflectingdata dependencies, and solving a coherent enterprise designproblem.

One simple example of a collaborative engineering de-sign problem is the Aircraft Concept Size (ACS) prob-lem [4]. This problem involves the preliminary sizing ofthe general aviation aircraft subject to certain performancecontraints. It consists of three disciplines: aerodynamics,weight, and performance. These three disciplines can berepresented as three optimization models with clearly de-fined input and output as shown in Figure 3. They can re-side at distributed locations. The enterprise analyst gives theconstraints and goals for the aircraft design. Three modelsneed to participate the optimization process with multipleinterations with the enterprise analyst.

Figure 2. Web Services Deployment for Dis-tributed Multilevel Design Optimization

3. Semantic Web Services Technologies

Fundamental web services technologies, such as WSDL,SOAP, and UDDI, provide e-Business a way to automateand streamline distributed business processes. The combi-nation of web services technologies and the semantic webcan facilitate the autonomous and heterogeneous applica-tions, data, and services residing in distributed environ-ments. In order to accomplish this goal, several issues areaddressed in semantic web services research.

� Service description. In order to enable automatic dis-covery, invocation, and composition of web services, acomputer interpretable description is needed. The pro-posed standards, such as DAML-S and OWL-S [10],provide an upper ontology for describing not only theinput and output but also properties and capabilities ofweb services.

� Service publishing and discovering. In order to letother people use the service, service providers shouldregister their services in a public registry or a pri-vate registry inside of an organization. A service con-sumer needs to discover a service that meets the func-tional requirements. UDDI [13] provides a standard-ized method for publishing and discovering informa-tion about web services. However, in order to supportthe automation of the discovery process, the registra-tion information should include both syntactic and se-mantic description of services.

� Service composition. Simple autonomous servicesare not enough to represent a large application process.

Figure 3. Aircraft Concept Size (ACS) problem(adapted from Gu and Renaud[4])

An approach is needed to generate a composite servicebased on existing ones. The composition process canbe done manually, semi-automatically [12], or auto-matically [2].

� Service execution. Service execution for compositionservices largely depends on the composition model.Either by transferring the composition model into aparticular workflow specification, such as BPEL4WS,and executed on a workflow engine [6], or by repre-senting the composing process using DAML-S specifi-cation and executing it by a DAML-S Virtual Machine[12].

Although there are various proposed semantic web ser-vices technologies, no widely accepted service representa-tion and deployment standard exists. In the next section,we present our proposed framework that adapts the currentexisting technologies to facilitate the enterprise wide engi-neering design process.

4. The Semantic Web Services Based Frame-work

From the functional perspective, the framework shouldprovide the following components: semantic descriptionof web services, services publishing and discovery, servicecomposition engine, and service execution engine. Figure4 shows the infrastructure of the framework. The frame-work participants can be separated into three types: serviceproviders, service consumers (engineering analyst or enter-prise analyst), and service composer (enterprise analyst).

Service providers are the entities that offer simple webservices. They are responsible for describing their web ser-

Figure 4. The Semantic Web Service BasedFramework Architecture

vice by providing values to the properties, and publishingtheir services. Service consumers may be end users or otherweb services that invoke a web service: either a simple orcomposite one. Service composers are responsible for spec-ifying composite services.

In this paper, an elementary service is defined as a simpleservice that does not interact with other services. A compos-ite service is defined as a complex process that is composedof multiple elementary services.

4.1. Semantic Description of Web Services

DAML-S and OWL-S are two emerging ontologies fordescribing the capabilities and properties of web services.They can be used to describe not only the elementary ser-vices but aslo the composite services. Medjahed et al [7]propose a semantic description method and composablitymodel for web services. Although the latter approach doesnot provide the semantic description of composite services,we discover that it is more flexible to add semantic descrip-tion for elementary services.

Therefore, We adapt the semantic description methodand composability model for web services that have beendescribed in [7] and use OWL [9] to specify the web ser-vices ontology. We model the ontology using a directedgraph. The web service ontology consists of WSDL con-cepts and extended features to represent the semantic capa-bilities.

Extending the approach of Medjahed et al [7], we addmore syntactic and semantic description for a web service,such as the URI of the WSDL file. Also for the qualitydefinition of web service, we add the availability and thelatency of execution the service.

OilEd [8] and Protege [11] are ontology and knowledge-

based visual editors. Jena [5] is an open source toolkit usedto create semantic web applications. It implements the RDFand OWL language and provides ontology API and infer-ence subsystems. The Jena toolkit can be used to build theknowledge base for web services. Both tools can be used tomanagement the ontology of web services.

4.2. Service Registration and Discovery

Elementary web services can be created with differentlanguages and deployed on various platforms, such as J2EEand .NET. Elementary web services can be separated intoseveral categories, including information input services, in-formation query and data analysis services, computationservices, and computing resource management services.

Service providers are responsible for creating and de-ploying these elementary web services. In the engineeringdesign domain, some elementary web services can be cre-ated by wrapping existing proprietary applications.

In addition to elementary web services creation, the ser-vice providers also need to register the syntactic and seman-tic information of their services. The service registrationprocess is shown in Figure 5. The provider needs to createthe WSDL, which defines the service’s syntax by specifyingthe structure of the input and output message, along with theservice’s runtime bindings.

WSDL

ServiceProvider

Agent

HTMLForm

1 OWL4

2

2

3

4

Database

Figure 5. Service Registration Process forSyntactic and Semantic information

1. The service provider starts the registration process byproviding the URI of the WSDL file.

2. A software agent on the service registration side ex-tracts the information from the WSDL file. Accordingto the WSDL information, a software agent generatesa HTML form that requires the semantic description ofservice and operations.

3. The services providers can fill out the form to specifythe meaning of the WSDL.

4. Then the software agent generates a OWL descriptionfor the service and stores it into the database.

Service consumers can discover a service by first identi-fing the required functionality. Then they access the servicecategory through taxonomies and find the approximate ac-curate service.

4.3. Service Composition

For a system with a small number of elementary web ser-vices, the domain experts can manually define a set of rulesdescribing the relationship among existing services and theapplication process logic. However, for a large scale systemthat consists of a large number of services and new emerg-ing services, manually predefining the composability rulesis not practicable. With the semantic description of webservices, it is possible to define the composability rules forcreating a composite service. Medjahed et al [7] identifytwo sets of composability rules to compare syntactic andsemantic properties of web services. We adopt four of theirrules for the service composition.

1. Mode composability compares the operation mode.

2. Binding composability checks the communication pro-tocol of interacting services.

3. Message composability compares the number of mes-sage parameters, data types, business roles, and units.

4. Operation semantics composability compares the se-mantics of service operations.

The workflow composition engine in the framework pro-vides the service composer a way to construct a compositeservice. It is intended to support three composition modes:ad hoc, semi-automated, and automated compositions.

� Ad Hoc composition: Under certain situations, theservice composer (enterprise analyst) knows the ap-plication logic and exact components that should beincluded in the application process. The service com-poser can form the composite service manually. Theworkflow composition engine needs to validate thecomposite service.

� Semi-automated composition: The workflow compo-sition engine lets the service composer create the com-position process by presenting all the optional servicesat each step. The end user makes the decision.

� Automated composition: The workflow compositionengine lets the service composer provide the input andoutput information. These information are fed into anAI planner, such as SAPA [3]. The planner returns oneplan, multiple plans, or no results to the end user forfurther decision.

If the service composer is satisfied with the generated work-flow, the workflow composition engine assigns an identifierto the workflow. The workflow specification, either rep-resented by an existing language (BPEL4WS for example)or a self-defined XML document, is generated to presentthe workflow for this task. The specification is stored in adatabase corresponding to the task identifier. Next, it is sentto the workflow execution engine.

4.4. Composite Service Execution

In the framework, the workflow execution engine con-sists of a central controller and a background processor [14].

� Central controller is implemented by exploiting JavaServlets technologies. It is responsible for handling thesynchronous web service invocation for short runningjobs such as data input that requires maintain the statusof interactions between end users and services.

� Background processor is responsible for handling theasynchronous invocation for long running jobs such asthe computation services. It monitors the status of allthe submitted tasks by checking the database periodi-cally. When it discovers that a workflow is ready forinvoking a computation service, it sends a SOAP mes-sage to the server.

This approach allows end users to submit their work-flow specifications, input the parameters interactivelywith services, and disconnect from the system and re-trieve all the data later. SAIWS [15], a simple asyn-chronous invocation framework for web services, isbuilt upon the Apache Axis that can be used for thispurpose.

There are two issues involved in the service executionstage need to be considered.

� Locating an alternative service

One of the issues in the service execution is that the selectedservice in the workflow is unavailable or off line temporar-ily. Then the execution engine invokes the alternative ser-vice if there is one defined in the workflow at the servicecomposition stage. Otherwise, the Service finder in the ex-ecution engine searches the knowledge base to find a alter-native service if there is one.

� Monitoring service execution

Web services often need long time to complete the execu-tion. Services consumers need to know what the status oftheir requesting services. For executing an elementary ser-vice, a service consumer may want to know if the serviceis completed at a time point or if an unexpected exception

occured. For executing a composite service, a service con-sumer needs to know where in the process the request is.Monitoring the service execution status is another impor-tant issue.

In this framework, the execution status of the web ser-vice is recorded in the database. The status of a web ser-vice includes: which requests are finished, which requestis invoked but not finished at this point, and if any unex-epected exception occured for the current invoked request.Service consumers can retrieve the executing status of theirrequested service from a web browser and decide if theywant to stop the execution. After the workflow executionis finished, the Email service sends an email to the serviceconsumer to notify of the completion.

5. Conclusion and Future Work

The presented framework is intended to support the E-engineering application, specifically the enterprise widecollaborative engineering design problem. However, it canbe easily adapted to support E-science applications, such asdistributed scientific simulations. The framework empha-sizes several issues for using semantic web services tech-nologies including semantic description of service, servicesregistration and discovery, services composition, and ser-vice execution.

More functionalities of web service execution monitor-ing need to be address in the framework in the future. Cur-rently, the framework allows end users to access the exe-cution status of their requested services. However, serviceconsumers can not change their plans after a composite ser-vice is invoked. Also at the current stage, the execution of acomposite service is centralized because it is easy to realize.However, a centralized execution model posses scalability,availability, and security problems [1] as the emerging peer-to-peer execution model becomes more attractive.

Acknowledgement. The authors would like to thankDr. John E. Renaud, who is a professor at the Departmentof Aerospace and Mechanical Engineering in University ofNotre Dame, for his discussions and input on the collabora-tive engineering design problem.

References

[1] B. Benatallah, M. Dumas, Q. Z. Sheng, and A. H. Ngu.Declarative composition and peer-to-peer provisioning ofdynamic web services. In Proceedings of the 18th Interna-tional Conference on Data Engineering (ICDE’02), 2002.

[2] M. Carman, L. Serafini, and P. Traverso. Web service com-position as planning. In ICAPS 2003 Workshop on Planningfor Web Services, 2003.

[3] M. Do and S. Kamghampati. Sapa: A multiobjective met-ric temporal planner. Journal of Artificial Intelligence Re-search, pages 155–194, 12 2003.

[4] X. Gu and J. E. Renaud. Implicit uncertainty propaga-tion for robust collaborative optimization. In Proceedingsof DETC’01 ASME 2001 Design Engineering TechnicalConference and Computers and Information in EngineeringConference, 2001.

[5] Jena. http://jena.sourceforge.net/index.html.[6] D. J. Mandell and S. A. McIlrith. Adapting BPEL4WS for

the semantic web: The bottom-up approach to web serviceinteroperation. In Second International Semantic Web Con-ference (ISWC2003), 2003.

[7] B. Medjahed, A. Bouguettaya, and A. K. Elmagarmid. Com-posing web services on the semantic web. The VLDB Jour-nal, 2003.

[8] Oiled. http://oiled.man.ac.uk.[9] OWL. http://www.w3.org/2001/sw/WebOnt/.

[10] OWL-S. http://www.daml.org/services/owl-s/1.0/.[11] The protege project. http://protege.stanford.edu/.[12] E. Sirin, J. A. Hendler, and V. Parsia. Semi-automatic com-

position of web services using semantic descriptions. In WebServices: Modeling, Architecture and Infrastructure work-shop in conjuction with ICEIS2003, 2003.

[13] Oasis UDDI. http://www.uddi.org.[14] X. Xiang and G. Madey. A semantic web services enabled

web portal architecture. In International Conference on WebServices (ICWS2004), 2004.

[15] U. Zdun, M. Voelter, and M. Kircher. Design and implemen-tation of an asynchronous invocation framework for web ser-vices. In The International Conference on Web Services -Europe 2003 (ICWS-Europe’03), 2003.