an integrated and collaborative approach for complex product development in distributed...

This article was downloaded by: [University of Colorado at Boulder Libraries]On: 20 December 2014, At: 07:45Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of ProductionResearchPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/tprs20

An integrated and collaborativeapproach for complex productdevelopment in distributedheterogeneous environmentH. W. Wang a & H. M. Zhang aa CIMS-ERC Tsinghua University , Beijing 100084, ChinaPublished online: 17 Mar 2008.

To cite this article: H. W. Wang & H. M. Zhang (2008) An integrated and collaborative approach forcomplex product development in distributed heterogeneous environment, International Journal ofProduction Research, 46:9, 2345-2361, DOI: 10.1080/00207540701738193

To link to this article: http://dx.doi.org/10.1080/00207540701738193

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/loi/tprs20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/00207540701738193

http://dx.doi.org/10.1080/00207540701738193

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

International Journal of Production Research,Vol. 46, No. 9, 1 May 2008, 2345–2361

An integrated and collaborative approach for complex product

development in distributed heterogeneous environment

H. W. WANG and H. M. ZHANG*

CIMS-ERC Tsinghua University, Beijing 100084, China

(Revision received October 2007)

Multidisciplinary collaborative modeling and simulation is an effective approachto verify the dynamic behaviour of a complex product such as an aerospacevehicle or a locomotive. The design environment of such products must cater forthe requirement for geographical distribution and structural heterogeneity in themodern industrial context. The focus of our research is a solution for such acollaborative modeling and simulation environment. Our approach is to utilizethe IEEE standard High Level Architecture for distributed simultaneoussimulation execution and management. At the same time, Web servicestechnology is employed to deploy and integrate functional models for thesimulation. The method for model description in this framework is proposed. Thetechniques for simulation execution and service encapsulation are discussed.A prototype system is implemented and a case study of locomotive tiltingcomponent design is executed. We find that both High Level Architecture andWeb services have their expertise for collaborative simulation and the integrationof these two technologies does achieve better interoperability and reusabilityamong heterogeneous simulation components in a distributed environment. Thenew approach has considerable promise for the efficient and successful complexproduct design.

Keywords: Collaborative modeling and simulation; High Level Architecture; Webservices; Complex product design

1. Introduction

To meet the requirement of military application, advanced distributed Modeling andSimulation (M&S) technology is proposed and is under constant development.Depending upon the robustness of the signaling messages transmitted across thenetwork and the evolvement of distributed M&S architecture, it is feasible to employthis technology in engineering and manufacturing processes in modern industry.

High Level Architecture (HLA) was initiated by the US Defense Modeling andSimulation Office (DMSO) in 1996. It, as defined by IEEE 1516, has many of theconstructive capabilities of the Aggregate Level Simulation Protocol (ALSP) plus thevirtual and live capabilities of Distributed Interactive Simulation (DIS). HLAenables the interoperability among diverse simulation systems and the reusabilityof legacy models and codes. The paradigm of Component-based Simulation (CBS)

*Corresponding author. Email: [email protected]

International Journal of Production Research

ISSN 0020–7543 print/ISSN 1366–588X online � 2008 Taylor & Francis

http://www.tandf.co.uk/journals

DOI: 10.1080/00207540701738193

Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

is achieved through the object model template (OMT) of HLA and the datainteraction and synchronization of simulation components are guaranteed throughthe HLA specification (Shen et al. 2002).

With the development of emerging Web standards and Web service models,great emphasis has been put upon achieving interoperability among previouslyindependent software elements utilizing these standards. Web services technologyprovides open interfaces for applications and enables the automated integration ofservices. It is a new paradigm for Web application and becomes the focus ofresearch for its promise in the integration of business process. Web servicestechnology is also attractive to the simulation world. US DMSO initiatedExtensible Modeling and Simulation Framework (XMSF) to encourage theapplication of Web technologies and open standards by M&S planners, managers,developers and users in April 2002. XMSF is defined as a modeling-&-simulation-tailored set of self-consistent standards, processes and practices employing a set ofWeb-based technologies and services to enable a new generation of internet-distributed applications to emerge, develop and interoperate. XMSF provides thetechnical basis for transformational interoperability via Extensible MarkupLanguage (XML) interchange, profiles, and recommended practices for Web-based modeling and simulation. It provides support for all types and domains ofmodeling and simulation (constructive, live, virtual, and analytical) (Pullen et al.2005, Sohn 2005). Web technology and Web services technology is also employedin civil application. The Web-based simulation system is studied (Miller et al.2000). The concepts of Model Driven Architecture (MDA) and Service OrientedArchitecture (SOA) are valued by scholars in underpinning the interoperability ofapplication in industry (Jardim-Goncalves et al. 2006). Simulation Web services areconstructed with the advanced distributed computing technology (Kilgore 2002).Process-centric engineering Web services are implemented in a distributed andcollaborative environment (Lee et al. 2006).

A complex product usually consists of multidisciplinary parameters andconstraints. Consequently, the design and development of complex product ismultidisciplinary in nature. It is necessary to consider the design problem globallyand taking into account various aspects. To analyse a complex product accurately,not only all the contents of a single discipline, but also the interactions andcooperation among different disciplines should be considered. The nature of complexproduct design desires a synthetic approach to evaluate the design. The virtualprototyping (VP) is a digital design method and technology, which is based on M&Sof various engineering disciplines and reflects the characters of the complex productfrom different perspectives in the early stage of design. The typical development ofmultidisciplinary collaborative simulation is the Vehicle Attitude Control (VAC)system of Ford.

Realizing the merits of the Web technology and HLA, researchers and developershave been actively exploring the application of these technologies in design andmanufacturing systems. A Web-based system framework for monitoring andcontrolling multidisciplinary design projects is described (Rogers and Salas 1999).A Web-based conceptual design framework is proposed and the prototype system isconstructed (Qin et al. 2003). A Web-based process planning optimization systemhas been developed to support distributed design (Li et al. 2005). Web servicestechnology begins to be applied to wrap legacy simulation modules (Tsai et al. 2006).

2346 H. W. Wang and H. M. Zhang

Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

Web services-enabled implementation of HLA is released for wide area simulationexecution (Pullen et al. 2005). Research on the integration of HLA with openWeb standard has also been carried out (Johnson 2002, Shen et al. 2002). However,Web-based systems are inevitably less effective for synchronized simulationexecution without the support of distributed simulation standards. The implementa-tion of an HLA standard through Web services still fails to solve the problem born ofthe standard itself.

Our solution for collaborative product development is to leave the wrapping ofthe model and the management of simulation uncoupled, which means the two partsare implemented as separate modules. We design a framework leveraging HLA andWeb services technology, in which Web services technology is utilized to provideremote communication while HLA is primarily for simulation traffic. Furthermore,we propose approaches feasible for our framework such as model processingmechanisms, service encapsulation and simulation execution management. Theconsistency of modeling and simulation is guaranteed through a model descriptionmethod and software modules are developed for seamless integration of the wholeprocess. A case study of locomotive tilting component simulation is executed as anillustration for the approaches.

The remainder of this paper is structured as follows. Section 2 presents anHLA-based collaborative simulation with the background of complex productdesign. Section 3 discusses Web services technology and its application in thesimulation domain. Also, a method for service encapsulation is proposed.A framework for integration of HLA and Web services technology is outlined insection 4. Finally, a case study of locomotive tilting component development isintroduced in section 5.

2. HLA-based collaborative modeling and simulation

HLA is initiated to promote the interoperability between diverse simulations andto improve the reusability of legacy models. HLA consists of three parts: HLArules, interface specification and OMT. HLA attempts to specify a standardinterface between simulations and separates these interfaces from the implementof any specific simulation. Within HLA, federations are defined as a group offederates forming a community. The formation of federations is a number ofsimulation components named federates that exchange information in the formof objects or interactions. The possible messages among federates of a federationare defined as a Federation Object Model (FOM). The capabilities of a federateto interact with others are defined as a Simulation Object Model (SOM) (Shenet al. 2002).

2.1 Modeling and simulation process based on HLA

The guidelines for designing and building an HLA federation are initiated as aFederation Development and Execution Process (FEDEP), which provides a genericprocess for simulation practitioners to construct standard HLA-based simulationfederations.

Collaborative approach for complex product development 2347

Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

The process for collaborative modeling and simulation development for complexproduct design is generally defined as follows:

(1) Design. Analyse the conceptual requirement of the system. Divide the systeminto sub-models according to their functionality. Treat these sub-models asblack boxes with only the input and output (I/O) data and time advancementto concern. Draft a description schema for the models and their I/O data.

(2) Development. Develop the FOM and SOM of the federation according to thedescription schema. Engineers start to design models belonging to a specificdiscipline. Choose an adaptor as the model transformation middleware.

(3) Testing. Execute the collaborative simulation on the basis of multidisciplinarymodels transformed by adaptor collaboratively.

(4) Analysis. Analyse the result of the simulation and collect enough informationto find potential problems in the design. Refine the design according to thedynamic behaviour of the design object.

Generally speaking, the models in complex product design are constructed incommercial computer aided tools. These models have their mechanism for I/Oprocessing and specific mechanism of time advancement. Here, we employ theconcept of an adaptor to make these models compatible with HLA simulationfederation protocol.

2.2 Model transformation

All possible messages in exchange in the federation are called FOM. FOM is thebackbone for interoperability in HLA. The rules of HLA restrict federates to sharingFOM data via Run-time infrastructure (RTI, the software implementation of HLA).Thus, the mapping and interoperation between the model data and FOM data in thesimulation federation are necessary.

2.2.1 HLA adaptor. The mechanism of model transformation is based on the HLAadaptor component that wraps the disciplinary model. Real-time message interactionis requested and reflected through RTI service and RTI callback. This mechanism iscontrolled by RTI through the paradigms of RTIAmbassador andFederateAmbassador.

There are principally two parts in the adaptor component: the modeltransformation facility and the simulation time advancement facility, as shownin figure 1. The model transformation facility consists of three parts: modeloperation, data interface mapping and interaction maintenance. Model operation isa low level operation for the disciplinary model which is concerned with specificsoftware utilized to construct the model. The model operation facility wraps thebehaviours of commercial software (start, pause, simulate to a specific time, stop,etc.). Data interface mapping is proposed to realize the high level operation ofdisciplinary models. Here, high level means caring nothing about the details of aspecific model and the corresponding commercial software. The mapping operationis between the variables in the disciplinary model and the SOM data of the federatemodel. Interaction maintenance is a mechanism for processing interactions properly.Since the RTI callback can happen at any time, the interaction maintenance facility


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

deals with the callback in a multi-threaded way and stores the interactions as eventsin an event queue. When it is driven by simulation step or event, the adaptor de-queues the interactions and deals with them for message dissemination or simulationtime advancement.

2.2.2 Mechanism of time advancement. The mechanism of time advancement is ofvital importance to distributed simulation because of the latency of the network.Time management rules in HLA facilitate reuse of simulation components because awide range of internal time management mechanisms of the federate are supported.At the same time, time management rules promote interoperability by allowingsimulations with different internal time management mechanisms to be combined ina single federation execution. Three types of time advancement mechanism aresupported in HLA: event-driven, time-stepped and the optimistic time advancement.In collaborative simulation of complex product design, time-stepped way of timeadvance is a good choice. First, the interaction of data is of more concern in suchsimulations; internal event rarely happens. Secondly, the disciplinary models oftenadvance internal time by small steps. Consequently it is possible to get the sameamount of results in each step of federate time which will improve the precision ofsimulation. Furthermore, observing the behaviour of the system as the advancementof time is more comprehensible to people.

2.2.3 Step mismatch. In the collaborative simulation environment, disciplinarymodels are often constructed through commercial software whose advancement stepmay be rather small to guarantee the precision of the simulation. Consequently, thenetwork charge is very high, which must be considered in the design of the adaptortime advancement facility. Our approach to this problem is to update data to RTI ina several-step interval. However, a step mismatch may happen in this approach. Asshown in figure 2, the adaptor stops at discrete time of the lower time axis whiledisciplinary models only advance to discrete time of the upper time axis.

Figure 1. Structure of the HLA adaptor.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

A mechanism of interpolation could be employed to improve the precision

lowered by step mismatch. For n sample points, the interpolation formula is defined

as follows:

IðTÞ ¼Xni¼1

liðTÞ �Oi ð1Þ

where T is the time variable of the federate; ti is the temporal points when the

disciplinary model generates output data; Oi is the output of the disciplinary model

at time ti; and I(T) is the interpolated input for the federate model.The coefficients in (1) can be computed in the equations below:

liðTÞ ¼ fiðTÞ=giðTÞ i ¼ 1, 2 . . . n

fiðTÞ ¼Qn

j¼1,j6¼i

ðT� tjÞ i ¼ 1, 2 . . . n

giðTÞ ¼Qn

j¼1,j6¼i

ðti � tjÞ i ¼ 1, 2 . . . n

8>>>>><>>>>>:

ð2Þ

2.2.4 Signal reconstruction. Some disciplinary models may require the input signal

in the form of a continuous temporal sequence. However, the federate model that is

responsible for the input of the disciplinary model always gets simulation data in the

way of interaction or object attribute reflections. Consequently, a mismatch between

the federate model output data and the disciplinary model input data occurs. An

approach of signal reconstruction could be employed to solve it in two ways. First, to

reconstruct the data through a one-order hold operation. Secondly, to apply the

interpolation and extrapolation in the disciplinary model if it is possible. The

formula of the one-order hold operation is shown as follows:

VðtÞ ¼ ðVTþfederate step � VTÞ � ðt� TÞ=ðfederate stepÞ ð3Þ

where VT is the value that the federate gets at federate time T; VTþfederate_step is the

value that the federate gets at federate time Tþ federate_step; V(t) is the value after

a one-order hold operation; t is the continuous time for disciplinary model after

one-order hold operation; and federate_step is the time step of the federate.The formation of the input data for the disciplinary model is generally an array of

temporal values, so the continuous value reconstructed from one-order hold

operation is again transformed to discrete value through sampling with a rather

small sample interval in the adaptor code.

Figure 2. Step mismatch in time advancement.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

2.3 Other issues

The HLA adaptor facility is able to solve problems in bridging the disciplinary modeland HLA standard. However, problems with the disciplinary model or themanagement of simulation still happen during our development work. Some issuesare addressed here.

Some software packages have no mechanism for recording the values of the lastsimulation when a new simulation step starts. Consequently, if there are integraloperations in the disciplinary model, the result will not be correct since the integralstate is lost. So, whenever it is necessary to record the internal state, an approach ofdefining corresponding variables for recording the state should be adopted. At thesame time, the Management Object Model (MOM) of HLA provides partialexecution management. However, an expansion of MOM through defining severalinteraction classes is sensible. For example, when internal errors occur, the federatecan identify the error through the interaction class. At the same time, OwnershipManagement (OM) in HLA provides a mechanism for attributing ownershipacquisition, but there is no callback for acquisition failure. Thus, an interaction classcan be used here. The agile expansion of MOM will be very helpful for some specialrequirements of the simulation.

3. Web service and encapsulation

As an emerging information technology which enables cross application integration,Web services technology is widely advocated in the industrial world. At the sametime, commercial implementations of Web services facilitate the encapsulation task.

3.1 Web service overview

Web service is a self-contained, self-describing unit of modularity for publishing anddelivering XML-based digital services over the Internet. Web services are consumedby accepting and returning messages (encoded in XML). A Web service’s externalbehaviour is described in terms of the syntax, semantics, and sequencing of messagesexchanged through an XML schema vocabulary. The agreed upon standards in Webservice are XML, Simple Object Access Protocol (SOAP) for messaging, Web servicedescription language (WSDL) for describing and Universal Description, Discoveryand Integration (UDDI) for integration.

XML is regarded as the meta-language of the Web. It is perceived as the mediafor heterogeneous system information exchange. XML is already being used fordata, content, messaging, and computing to provide point-to-point integration in aplatform-neutral way. XML schemas provide a standard way to define the structure,content and semantics for XML documents.

SOAP is an XML-based, lightweight messaging protocol for exchange of typedinformation in decentralized, distributed environments (Sohn 2005). SOAP simplyconsists of a modular packaging mechanism and a set of encoding rules. Therefore,it does not depend on any specific object model or programming language. Theplatform-neutral encoding mechanism enables applications on different platformsto exchange information. A SOAP message is a one-way transmission routed along


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

a message path. Each SOAP payload is an XML document with the root element5Envelope4 that wraps all of the specific elements and attributes. Within the5Envelope4 is an optional5Header4 for protocol extensions and the mandatory5Body4 element. The 5Body4 will contain one distinct root element that willidentify it as a request or response object. SOAP includes a built-in set of rules forencoding data types which enables the SOAP message to indicate specific data types,such as integer, double or arrays. SOAP data types are divided into two broadcategories: scalar types and compound types. Also an entire XML document canbe embedded in a SOAP message ignoring encoding style, which is calledliteral encoding.

3.2 Web services in simulation

The advocator of Web services in simulation is XMSF, which is dedicated topromoting a new generation internet distributed simulation. XMSF hopes to realizethe integration of heterogeneous simulation systems such as the HLA basesimulation and traditional Command Control Communication Computer& Intelligence (C4I) system through the standard Web services.

Practitioners of XMSF have created a strong technical relationship with keystandard organizations, e.g. HLA technical support team, Object ManagementGroup (OMG), Simulation Interoperability Standards Organization (SISO),Web-enabled Simulation Consortium (WebSim),W3C, etc. (Blais et al. 2005). Tocreate practical understanding of the application of XMSF precepts to real products,SISO established an XMSF Profiles Study Group in September 2003. The group isactively defining formal technical specifications for application of interoperableWeb-based technologies enabling composable and reusable M&S elements, andfacilitating enterprise integration. The group points out that the specification ofXMSF should be a collection of profiles detailing how to interoperate with XMSFcompliant systems. The group also defines the micro level of and objects of a profilelike Web-enabled RTI (Pullen et al. 2005). Much practical researches have beencarried out to integrate existing software via Web services.

XMSF is in its fourth year and has proved to be successful in methodology andtechnology. To bridge the gap that arises when different simulations are supportedby diverse infrastructures, XMSF applies a Web services and grid computingapproach in the technical level, while at the conceptual level advocates the techniquesfrom MDA. Web services technology is the technical backbone for integration ofsimulation systems, while MDA helps to realize meaningful integration. HLA isinherited by XMSF and cooperates with Web services and grid computing to provideinteroperability between highly distributed systems and models with differentgranularity (Pullen et al. 2005). XMSF is sponsored by the military and extends itsconception to civil simulation application. Many practical applications of XMSFprove that open standards such as the Web service and grid are exciting things in thesimulation world.

3.3 Service encapsulation

There are many protocols for implementing Web service, for example FTP,Hypertext Transfer Protocol (HTTP), SMTP. In the XMSF profile Web-enabled


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

RTI, Blocks Extensible Exchange Protocol (BEEP) and SOAP are applied. BEEP

allows bi-directional calls through the interface which enables RTI service calls and

call-backs (Morse et al. 2004). The service-oriented distributed computing

environment provides a powerful approach for integration of enterprise resources

and legacy models. A lot of effort has been made to publish legacy resources as

services (Kilgore 2002).In our opinion, what is sensible is to use commercial realizations and integrate

them in enterprise from a scalability and openness perspective. Our approach for

service encapsulation is based on the HTTP implementation of ApacheTM Software

foundation.

3.3.1 Structure of service encapsulation. The rationale for encapsulation is shown infigure 3. There are four parts in our HTTP-based Web service encapsulation:

(1) A Web server. This is the container for the HTTP requests, responses and

SOAP engine.(2) SOAP engine. This provides a mechanism for sending and receiving SOAP

messages, XML serialization and de-serialization.(3) Web service adaptor. This is the interface between the legacy resource and

SOAP engine. It receives data from legacy resource and serializes it in XML

stream format. It de-serializes the XML stream and pushes the data to legacy

resource.(4) Legacy resource. This consists of software modules and simulation systems

that should be reused in collaborative simulation.

The SOAP engine is a package deployed in the Tomcat Web server as a Web

application. The key facility for the SOAP engine is a servlet that deals with HTTP

request and response. The SOAP engine provides a series of API for object level

encapsulation in the form of XML format messages. The Web service adaptor

functions like the HLA adaptor, which is initialized in the deployed code that

embodies the service. The Web service adaptor translates formatted messages

between the SOAP engine and the legacy resource.

Figure 3. Structure of service encapsulation.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

3.3.2 Web service adaptor. Web service adaptor is a middleware between the SOAPengine and legacy resources. Consequently, it must implement the bi-directionalinterface to SOAP engine and legacy resources. Java plays an important role in thiscase. First Java can easily get access to many Java-based Web services. Secondly,Java provides a number of mechanisms for integration with non-Java software. Javaalso supports a wide range of network connection such as sockets, CORBA, RMI,etc. These facilities can be applied as solutions for connecting to other distributedcomputing resources. Furthermore, Java language enables access to the nativemethod through Java Native Interface (JNI) and has the platform independent andeasily implemented features, which make it a better choice for system implementa-tion among diverse network and workstations. Web service adaptor exchanges datawith SOAP engine through serialization and de-serialization. It provides severalseparate facilities that make the seamless integration of the legacy resource and Webservice possible. The resource description schema describes the dynamic characters ofa Web service such as the input data and output data. The profile paradigm providesa mechanism for accessing legacy resource. The profile describes a group of modules,packages, interfaces to get access to the disciplinary model constructed in MatlabTM,AdamsTM, etc. Moreover, the interpolation and one-order hold operation that areintroduced in HLA adaptor still work well with the Web service adaptor. Based onthe resource description schema, variable mapping and data validation are enabled.Variable mapping is to make a local reference in the adaptor for disciplinaryvariable, so the variables to access are dynamically defined. At the same time, thedescription schema defines the input and output spans which make data validationfeasible. The Web service adaptor in our implementation is designed to adaptable tovarious Web service and models to be encapsulated through dividing it into submodules as shown in figure 3. For example, if the SOAP engine is changed, justreplace the variable mapping module while other parts of the adaptor remainunchanged.

4. A HLA and Web service-based framework

As introduced in former sections. HLA is defined as the architecture with standardspecification and implementation. HLA aims to promote the reusability andinteroperability of legacy simulation packages towards a component-based simula-tion solution. Any federate who has the knowledge of FOM of a federation can joinin the simulation at any run time of the simulation. The reconstruction ofdisciplinary model through HLA adaptor paradigm in complex product designenables the disciplinary models to work in a federation and interact with each other.However, HLA-based collaborative simulation falls down in resource distribution.Moreover, drawbacks exist in the object model of HLA.

At the same time, models in engineering design can be encapsulated and deployedas Web services. Web services technology is based on open standard and can beaccessed in a highly internet-distributed way. Through Web service adaptor, legacyresources can be deployed as Web service; the powerful computing functionality canbe invoked remotely and the data is transported in firewall safe HTTP requests andresponses. However, Web services technology is limited in the management ofsimulation. Lots of problems exist in the synchronization of simulation and


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

ownership management of simulation data. The integration of HLA and Webservices will tremendously complement the drawbacks of these two infrastructures.

4.1 Framework of integration

In our framework of integration, Web-based technology is used to provide theremote communication with the user clients, while HLA is primarily for thesimulation traffic. The detailed structure of the framework is shown in figure 4.There are principally three parts in the framework: the part for collaborativemodeling, the part for Web service encapsulation and the part for HLA simulationmanagement.

The pre-simulation process of collaborative modeling is an important part. Thecollaborative modeling tool describes models in a four-view way which will beintroduced in section 4.2. The four-view description of the collaborative model ismapped to an XML schema which is called Collaborative Model DescriptionLanguage (CMDL) by us. Disciplinary models are encapsulated and deployed asWeb services geographically distributed across the enterprise. In local HLA analysisfederation, federate agents are constructed and serve as two roles, the first is theconsumer of Web service, and the second is a federate of the simulation federation.Federate agents call services from both Web service and RTI service. It has an enginewhich finishes the work of model parsing and mapping between Web service datatypes and RTI data types. CMDL is the guidance for Web service encapsulation andthe construction of local federate agent.

4.2 Four-view description of the collaborative model

The tool we design for collaborative modeling consists of four views for thecollaborative model. After defining the four views through a graphical user interface

Figure 4. Integration framework of HLA and Web services.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

(GUI), an XML schema describing the model is generated. It describes necessaryinformation about the simulation federation in HLA and the Web serviceencapsulation.

(1) The FOM view. The FOM view describes the FOM for the simulationfederation. Objects and interactions are defined in FOM view. Thus, thefederation description document (FDD) can be generated from the FOMview.

(2) The Service view. The service view is concerned with services that are used incollaborative simulation. Service description and deployment tool parses themodel information of service view and deploys the models as correspondingservices.

(3) The Actor and Mapping view. The functionality of this view is to define howmany federates are in the simulation and how the interaction of thesefederates are designed. The mapping is between the service view and the FOMview. Through the mapping operation, the SOM data for HLA is generatedand federate agents can be constructed.

(4) The Deployment view. It is the draft for the deployment of the wholecollaborative simulation system. First, the addresses and functions of Webservice are defined. Secondly, where the RTI runs and where federate agentsruns are also defined. Furthermore, the relationship of invoking betweenfederate agents and Web services are represented in this view.

The four-view representation is constructed in the supporting tools and resultsin an XML schema of CMDL for a full description of the collaborative simulation.A framework like this has the benefits of resource distribution, open standard based,scalability and robustness, etc. Difficulties lie in the development based on twocomplex infrastructures and the complex mappings. However, a framework like thisdoes make sense in designing an integrated collaborative simulation system.

5. A case study of locomotive development

In former sections, we discussed rationale for HLA-based collaborative M&S,Web service paradigm and service encapsulation. We also propose a frameworkbased on HLA and Web service for collaborative simulation in complex productdesign. On the basis of the conceptual design of our framework, we have developedenabling tools that facilitates collaborative simulation. The tools are thecollaborative modeling tool, service description and deployment tool, processmanagement tool for collaborative simulation and the simulation management tool.The GUI of service description and deployment tool and simulation managementtool are shown in figure 5. In this section, we will introduce a case study oflocomotive tilting component development.

The development of the locomotive is seen as a key technology for the successfulimplementation of the high-speed railway network. The development of thelocomotive has the characteristic of consisting of multidisciplinary techniques fromvarious engineering domains. In order to consider the design problem globally atthe early design stage, it is necessary to employ the collaborative modeling andsimulation technology.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

Figure 5. (a) Tool for service description and deployment. (b) Tool for simulationmanagement.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

In this case, the aim of the collaborative simulation is to get the dynamicresponding data when the locomotive moves and tilts with different velocities oncurved railway of different radii. The design of the locomotive multi-body dynamicssystem depends on the control model. The control model is responsible to drive thehydraulic module. The federation model consists of three disciplines principally, themechanical multi-body dynamics model, the control model and the hydraulic servomodel of the locomotive.

The mechanical system of the locomotive is built in AdamsTM, the control systemis built in MatlabTM and the hydraulic servo system model is built in HopsanTM. Wetry to encapsulate these models as Web services and access the functionality of thedistributed model. Thus, engineers can see the property of locomotive tiltingcomponent globally at the early stage of design. At the same time, engineers can turnto historical design models that can also be published as Web services. Consequently,engineers are able to reuse legacy resources and assemble legacy components withthis creative design.

The collaborative simulation is carried out in a procedure as follows: afterdefining the collaborative model through the modeling tool, the CMDL is generatedand disseminated to modeling engineers. Modeling engineers construct disciplinarymodels while simulation analysts and software engineers start the work of federationconstruction and service encapsulation. At last, all the models are deployedaccording to the deployment view of CMDL. Then a collaborative simulation isready to be executed. The result of the collaborative simulation is shown in figure 6.

Through collaborative simulation based on HLA and encapsulated services,multi-disciplinary models run collaboratively and generate a real-time result of thelocomotive tilting status. Depending upon the result of simulation analysis, engineersrefine the design and redeploy the model until the result caters with the requirement.Thus, collaboration between the participants is enhanced in this context; reusabilityand interoperability of enterprise resources are also achieved.

Comparing with the approaches as reviewed in the literatures, our collaborativeframework has the following advantages and disadvantages:

(1) Our solution is simulation and synthesis oriented. The framework is proposedto underpin the collaborative modeling and simulation so as to solve theproblems in the early stage of design process. It makes the collaborative workbetween product developers possible.

(2) The proposed framework is based on the matured HLA andWeb services. Thelegacy HLA-based simulation systems and codes could be easily reused. Thedevelopment work could be based on matured commercial products of HLAand Web services implementation with the exemption of getting into technicaldetails about HLA and Web services. It is very sensible for enterprises.

(3) The application is internet distributed. The collaborative work is extended toa wide area network. It is beneficial for the large scale corporations. Themodel and design expertise is published as Web services and could be reusedby any design work.

(4) The complexity of our framework is increased by the complexity of HLA andWeb services.

(5) There is some limitation in supporting the reuse of the whole federationmodel, which will be the key issue in our future work.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

Figure 6. (a) The dynamics model. (b) The tilting curve of the locomotive body.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

6. Conclusion

HLA and open Web technologies are powerful tools for modern M&S. A frameworkwhich leverages Web services and HLA is feasible from both conceptual andtechnical perspectives. A highly distributed collaborative M&S environment basedon this framework enables the interoperability between single disciplinary modelsand reusability of enterprise legacy models. Engineers can consider the designglobally at the early stage of design and reuse legacy models in their workthrough collaborative simulation in such an environment. Furthermore, it providesa uniform way of M&S management through enabling tools which tremendouslyimprove the efficiency of design. Consequently, a system like this does haveconsiderable promise for the simulation analysis of complex product design. In ourfuture work, emphasis will be put upon refining the description method of thecollaborative model, implementing the integration of collaborative simulation withenterprise information system and constructing Web-based collaborative designenvironment.

Acknowledgements

This paper is supported by the National Natural Science Foundation of China(grant No. 60674079) and the National R&D High-Tech Plan (863 Program)of China.

References

Blais, C., Brutzman, D. and Drake, D., Extensible modeling and simulation framework(XMSF) 2004 project summary report, Naval postgraduate school, MontereyCalifornia, February 2005.

Jardim-Goncalves, R., Grilo, A. and Steiger-Garcao, A., Challenging the interoperabilitybetween computers in industry with MDA and SOA. Comput. Ind., 2006, 57(8/9),679–689.

Johnson, G.D., Networked simulation with HLA and MODSIM III. Proceedings of theWinter Simulation Conference, 5–8 Dec., 2002, pp. 1065–1070.

Kilgore, R.A., Simulation web services with .NET technologies. Proceedings of the WinterSimulation Conference, 8–11 Dec., 2002, pp. 841–846.

Lee, J.Y., Lee, S., Kim, K., Kim, H. and Kim, C.-H., Process-centric engineering Web servicesin a distributed and collaborative environment. Comput. Ind. Eng, 2006, 51(2), 297–308.

Li, W.D., Ong, S.K. and Nee, A.Y.C., A Web-based process planning optimization system fordistributed design. Comput.-Aided Des., 2005, 37(9), 921–930.

Miller, J.A., Seila, A.F., Xiang, X., The JSIM web-based simulation environment. FutureGeneration Comput. Sys., 2000, 17, 119–133.

Morse, K., Drake, D. and Brunton, R., Web enabling HLA compliant simulations to supportnetwork centric applications. Proceedings of the 2004 Symposium on Command andControl Research and Technology, San Diego, CA, 2004.

Pullen, J.M., Brunton, R., Brutzman, D., Drake, D., Hieb, M., Morse, K.L. and Tolk, A.,Using Web services to integrate heterogeneous simulations in a grid environment.Future Generation Comput. Sys., 2005, 21(1), 97–106.

Qin, S.F., Harrison, R., West, A.A., Jordanov, I.N. and Wright, D.K., A framework ofweb-based conceptual design. Comput. Ind., 2003, 50(2), 153–164.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

Rogers, J.L. and Salas, A.O., Toward a more flexible Web-based framework formultidisciplinary design. Adv. Eng. Software, 1999, 30(7), 439–444.

Shen, X., Radakrishnan, T. and Georganas, N.D., vCOM: Electronic commerce in acollaborative virtual world. Electronic Commerce Res. Appl., 2002, 1(3/4), 281–300.

Sohn, M.M., Advanced M&S framework based on autonomous Web service in Korea:intelligent-XMSF approach. In Lecture Notes in Computer Science (3398), pp. 150–158,2005 (Springer: Berlin).

Tsai, W.T., Fan, Chun, Chen, Yinong and Ray, Paul, A service-oriented modeling andsimulation framework for rapid development of distributed applications. SimulationModelling Practice and Theory, 2006, 14(6), 725–739.


Dow

nloa

ded

by [

Uni

vers

ity o

f C

olor

ado

at B

ould

er L

ibra

ries

] at

07:

45 2

0 D

ecem

ber

2014

an integrated and collaborative approach for complex product development in distributed...

Documents