40. SBAI- SimpósioBrasileirode AutomaçãoInteligente, São Paulo, SP, 08-10 de Setembrode 1999

AGILE SUPPLV-CHAIN COORDINATION IN THE VIRTUAL ENTERPRISE '

Ricardo J. Rabelo, Alexandra P. Klen, Luiz M. Spinosa, Aureo C. FerreiraG-SIGMA - Intelligent Manufacturing Systems Group

UFSC / EMC / GRUCON - Cx. Postal 476CEP 88040-900 - Florianópolis - SC

Abstract: In this work the suppIy-chain co-ordination in thevirtual enterprise environment is subject of deeper analysis.The system developed by UFSC - DBPMS - for themanagement of distributed business process is presented andsome basic concepts for its definition and development aredetailed. The DBPMS provides means for getting, analyzing,making available and managing the information from andabout a virtual enterprise , enabling the enterprises to make theirlogistics more efficiently by means of a integrated information-based supply-chain management

Keywords: Virtual Enterprise, Supply Chain Management.

1 INTRODUCTIONEven more ' the use of information technology is beingencouraged for sharing and exehanging information rightacross individuaIs and organízations. This trend can beunderstood as a consequence of a new strategy of eonductingbusiness, which is the concept of Virtual Enterprises (VE).Aceording te the terminology and definition used by theprojeet eonsortium (Prodnet, 1996; Camarinha-Matos &al.,1997)1,the paradigm of virtual enterprise is a temporaryalliance of enterprises that come together to share skills andresources in order to better respond to business opportunitiesand whose cooperation is supported by computer networks andadequate Information Technology (lT) tools and protocols.

In the manufacturing sector, the Virtual enterprise is most1ycomposed by small and medium-sized enterprises behaving assuppliers and having no definite relations , policies andimplications. Therefore, it is not difficult to perceive the degreeof complexity to manage this kind of value-chain as weil as toco-ordinate the logistics of a business process that isdistributed.

The PRODNET-ll (production Planning and Management inan Extended Enterprise) is a three-year (October/96 -September/99) international cooperation project supported bythe Európean Union and by the CNPq (lhe Brazilian Councilfor Technological Development and Research). The project. aims to design and develop an open platform and lhe adequate

1 CSIN (P), ESTEC (P), HERTEN (BR), Lichen Informatique (F),MIRALAGO (P), Uninova (P), Uníverslty of Amsterdam (NL),Universidade Federal de Santa Catarina (BR), and Universidade Novade Lisboa (P).

289

IT protocols and mechanisms to support industrial virtualenterprises with special focus on the needs of smail andmedium sized enterprises (SMEs). PRODNET-ll focuses onmeans to inter-operate with severaI value chain networksemploying new emerging standards and advanced technologiesin communication, cooperative information management anddistributed decision making.

As a general requirement for an infrastructure to support VEs,it can be pointed out thatthe companies must be able to inter-operate and exchange information in real time so that they eanwork as a single integrated unit although keeping theirindepcndence / autonomy. It also has to be taken into accountthat legacy systems were not designed with lhe idea of directlyconnecting to corresponding systems in other enterprises.Typically, enterprises pre-exist before they decide to join in aninformation sharing and exchange network. Consequently,every enterprise is autonomous, developed independently fromother enterprises and uses distinct information managementand control strategies that serve its own purposes. Thus lhesituation is extremely heterogeneous and requires adaptation ofexisting production planning and control systems (pPC) toelectronic linking.

To support this environment and coping with lhe legacysystems in enterprises, PRODNET-ll proposes, for each nodeof aVE . network, an architecture where, among others,Aâvanceâ Co-ordination Functionatities address a fewadvanced co-ordínatíon aspects: (i) Partners search andselection; and (ii) Integrated logistics decision support system.The logistics decision support system aims at providing lheenterpríse with "real-time" information about the current statusof the orders in the involved enterprises.

In this work, lhe supply-chain co-ordination in lhe VEenvironment will be subjeet of deeper analysis. The systemdeveloped by UFSC for lhe management of distributedbusiness process wilI be presented and some basic concepts forits definition and development wiil be detailed.

2 BUSINESS PROCESSES IN THEVIRTUAL ENTERPRISE ENVIRONMENT

Thc Virtual Enterprises must also deal with orders requests. Inthis case, these orders are named distributed business processes(DBP). A DBP is a dynarnic and temporary set of "orders"

40. SBAI- Simpósio Brasileiro de Automação Inteligente, São Paulo, SP, 08-10 de Setembro de 1999

(business processes - BP) which jointly gives rise to the end client-enterprise, whereas node-I, node-3 and node-5 are itsproduct of the VE. As the BPs are supposed to be performed by direct suppliers. Each node has some (sub-) BPs under itsseveral enterprises, the enterprise that triggered the format ion responsibility (and so the respective Enterprise Activities -of a given VE normalIy must co-ordinate their execution in EAs), placed into the dashed circle around them. Theyorder to avoid business chaos (Rabelo et al., 1996). In a VE rcpresent the value added by each enterprise on the product ionscenario the management of the value-chain is a complex task, chain, along the time . For examplc, node-l has to perform fourespecialIy when some degrees of co-ordination are envisaged interdependent BPs (BPI_lO, BPI_22, BP l_22A andto be supported. Figure I illustrates a DBP concept via its BPI_22B), whose result has to be sent to node-2 so that theproduction graph. The enterprise '2' (node 2) is currently the BP1_1 can be done .and so the whole production can be kept,

Figure 1: DBP concept

__I BP_2

end-producf 2

..time horizon

Since these BPs are supposed to be performed by severalenterprises, the client-enterprise must manage their execution.As such, it is also liable to unexpected events, like BP delay,BP or DBP cancellation/rnodification, changes on DPBpriorities as welI as to local communication deficiency andnetwork overloading or failure. When one of these problemsoccurs and the responsible enterprise cannot solve the problemlocally, it offends the DBP contract, and then causes a conflict.This conflict basically influences the DBP production dates(the planned delivery date) and can affect the whole productionchain, Because the components of a DBP are alI inter-related,an enterprise cannot (re)plan itself for its own benefits only,but it must also look for the benefits of the whole net,Consequently, a close cooperation has to exist both intra-enterprise and inter-enterprises in order to minimize the globalloss in the "DBP contract",

The complexity of the resolution of a conflict may vary quite a101. Because an enterprise usualIy plans with some temporalintervals between the BPs (just to absorb the unexpectedoccurrences), a re-planning ncgotiation may be relegated toadjust the planned delivery date either to the earliest deliverydate or the latest delivery date. This policy could be onesolution for a rapid contract re-negotiation and hence for a fastDBP plan re-establishment. However, depending on theproblem severity, the solution may involve a deeper - andpossibly complex - analysis. Many considerations andevaluations have to be taken into account, specialIy regardingthat an enterprise usuaIly does not have only one DBPcontracted, but plenty of them, which in turn can be indirectlyaffected by the problematic BP. Thus, the whole complexitymakes almost impossible a user or a system solve the problemindividually, Neither a single user has alI knowledge andenough capabilities (in terms of time and technicalbackground) nor a system has the natural human 'businessflexibility' and experience for good trade-off analysis /decisions.

290

In that sense, a balanced approach seems to be the suitablesolution to contemplate lhe enterprise with these two'knowledge sources' sirnultaneously, In other words, a DSSwhich can help and assist the user in a decision-making, In thiscontext, a DSS has to cope with lhe folIowing stages (O'Neill,1995): identify the problem situation; data acquisition andanalysis; determine causes of problem; define objectives;generate alternative solutions; comparing and evaluatingaltemative solutions; select (the 'best ') solution.

A framework for this DSS must provide the decision-makerwith selected inforrnation, describing both the competi tiveexternai environrnent and the way the enterprise is operating.Besides that, it must also provide well-established decision-making support models and lechniques.

In that direction, the irnplemented DBPMS (DistributedBusiness Processes Management System) offers an enterprisewith an environment that provides reliable and timelyinformat ion about the supply-chain and a support for rapiddecision-rnaking, two extremely important enabler aspects tosupport the enterprise agility and hence its cornpetitiveness.

3 AGILE COORDINATIONThe co-ordination is a subject of great importance for therealization of aVE. Considering that the network of enterprisesis formed due to the requirements of specific orders, there mustbe a way to co-ordinate these activities, In the Prodnet-IIproject this is made through "Advanced Co-ordinationFunctionalities" (ACF) . ACFs correspond to specializedsoftware modules being developed -to help solving specificproblems that require co-ordination actions within the VE.

The co-ordination functionalities are, in fact, a subset of thefunctional requirements for aVE. The Prodnet-II architectureand implemented platform comprises a number of semi-autonomous software modules that provides the enterprise with

40. SBAI- Simpósio Brasileiro de Automação Inteligente , São Paulo, SP. 08-10 de Setembro de 1999

several other functionaJities, such as: exchange of cornrnercialdata (using EDIlEDIFACf and WWW), exchange of technicalproduct data (STEP), quality related information exchange, VEinformation management, and incomplete and imprecise ordersmanagement. Ali these "services" are supported by other basicsoftware modules to guarantee the privacy and safetycornrnunication, distributed data management, architecture /services configuration and global co-ordination among themodules (Camarinha-Matos &al., 1998).

From the enterprise point of view, ali those functionalities areseen as a "black box" which offers VE services - lhe ProdnetPlatform. It corresponds to an upper software layer to beinstalled on lhe "top of" enterprises' legacy systems (theirinternal modules) so that the enterprises can operate in a VEscenario in a "transparent" and integrated way. In fact, twomain modules constitute this platform: the ProdnetCooperation Layer (PCL) and the ACFmodules.

The ACFs can be seen as high-Ievel services to be offered to anenterprise/users and that uses the core services from the PCL's-modules. The cornrnunication between the ACFs and the PCL

is supported by means of a specific (but open) API(Application Program Interface).

The DBPMS represents an ACF which aims to provide meansfor getting, analyzing, making available and managing lheinformation from and about aVE, enabJing the enterprises tomake their logistics more efficiently by means of a integratedinf0"!1ation-based supply-chain management. It was designedto act mainly in the operation phase of a VE life cycle (Spinosa&al., 1998b), comprising a bit of the creation/configuration anddissolution phases.

In fact, the main fundamentais of the implemented DBPMSfunctionalities refleet lhe consideration of the emergent andwider concept of logistics called "integrated logisticsmanagernent" (Christopher, 1994). In that sense, the DBPMSextends lhe intra-organizational logistics carried out by lhec1assical PPC systerns, with a higher-Ievel vision about theVE's logistics, giving rise to an inter-organiuuional logistics(Klen &al., 1998b). Its general approach is iIIustrated in theFigure 2.

Figure 2: DBPMS General Approach

PCL

ShopfloorVE Coordinalor

The whole process starts when the VE-Coordinator (the VEtrigger) wants to monitor, after a given VE is created, theprocessing of a certain BP at a given VE-Mcmber's site (aSupplier). During lhe VE creation phase a set of agreernents ismade, including the specification of a number of informationthat the VE-Member 's PPC should send to the VE-Coordinator(like order status, remaining process time, the amount of partsalready produced , etc.) as well as the periodicity this should bemade.

These specification are indicated in the "supervision clauses",an innovative information structure logically aggregated to therespective BP's contract (the "purchase order's contract" in thePPCs), which is filled up by the user via a specific graphicalinterface. Supervision c1auses are used to specify the VE-Coordinator and VE-Members' rights and duties in terms ofinformation access for monitoring purposes. In general, theyspecify what, how, when, where and how far a given set ofinformation about an order should be monitored from each VEsupplier.

Once a given VE is created, ali the information about it - andthe one required by the DBPMS - is stored into the DistributedInformation Management System (DIMS). The informationbeing monitored (specified in the rights ofthe VE-Coordinator)is sent and managed by DIMS - one of the PCL's modules -which acts as federated and distributed database (Afsarmanesh

291

ShopfloorVEMember

&al., 1998), providing a "transparently", secure and eneryptedinter-communication between the VE-Coordinator and everyVE-Member (Osorio &al., 1998). Depending on thecomplexity of the transaction, another PCL's module isinvoked to better coordinate the PCL's services workflowintemally .

In the Prodnet-Il approach, the VE-Members ' PPCs shouldobtain the set of duty information (in predefined cycles of time)from their shop floors and put them available in their respectiveDIMS, so that the VE-Coordinator' s DBPMS can have accessto them in order to reason about. The continuous arrows in thefigure 2 iIIustrate this flow.

In the other direction there are the rights of the VE-Member, orin other words, the duties of the VE-Coord inator. It is alsoimportant to endow the suppliers with reliable information. Inthis sense, the VE-Coordinator may have to inform the VE-Member about a set of information, previously agreed andspecified in the supervision c1auses / duties . The dashed arrowsin the figure 2 iIIustrate this flow. It means that a minimum setof DBPMS services should also be provided to the VE-Member side.

The implemented DBPMS functionalities are comprised intothree main blocks:

440. SBAI- SimpósioBrasileirode Automação Inteligente, SãoPaulo,SP, 08-10 de Setembro de 1999

THE OBPMS FUNCTIONALlTIES aims to offer an electronic way to get information from lhesuppliers so that the enterprises can constantly update theirproduction plans and schedules (Ràbelo &al. , 1997).

4.1 Supervision Clauses ConfigurationIt allows lhe user to interactively configure the supervisionclauses to be applied upon every VE-Member of a given VE. Itcomprises the mapping between lhe VE-Members' PPCscapabilities in terms of information gathering against theDBPMS needs, and also the set of (different) clauses at theleveis of requested / purchase orders, required items andproduction orders.

4.2 VE SupervisionOne of the main probJems the enterprises face nowadays is lhedifficulty for obtaining reliable information from the partnersabout the real status of some contracted BPs, which directlyaffect lhe DBP (lhe entire client order). This situation obligesthe enterpriscs to frequently re-plan their actions due to theoccurrence of unexpected events, representing in delays oranticipations. It means an increasing of lhe enterprise's globalcosts because a delay or anticipation of one supplier can affectlhe execution of other inter-related BPs. This functional block

4.3 Oecision Support System (OSS)The introduction of a decision-support system (DSS) that helpsan enterprise to evaluate and to decide in the presence of aconfIict is a key point for its efficient reaction. This kind ofsystem can play its role much better if it is fed with reliable andreal-time information from the supply-chain, which is providedby lhe functionality previously explained. It incorporatesanalytical procedures as a support for the VE (re)schedulingand its evaluation, including means for a trade-off analysis(Spinosa &al., 1998a) (Rabelo &al., 1998). This DSS assumesa strong interaction with the enterprises' end-users,

Figure 3 illustrates lhe implemented DBPMS architecture,involving seven main modules to which the user can interactwith:

Supervision Clauses Configuration: configuration of the rightsand duties of the VE-Coordinator and VE-Members of a givenVE;

DBP monitoring: real-time information gathering from lhesuppliers' PPCS (in terms of orders) ;

Figure 3: DBPMS Architecture and the Main FunctionalitiesDSPMS

Conflicl detection: detection and identification of unexpectedproblems during lhe orders processing;

Reaclive decision-making: support for a decision-makingaccording to lhe conflict detected;

Simulation & assessment of alternatives : support for lheselection of a simulated decision by the user;

VE Analyzer: has lhe objective to provide an intra-organizational analysis of lhe VE in operation as well as aninter-organizational analysis of lhe alternative solutions.

DBP Control: set of actions that should be carried out in orderto implement a selected (simulated) decision.

In fact, those functionalities represent a subset of a widerfunctional requirement analysis, which was supported by aspecific methodology (Klen &al., 1998a) and that had threemain inputs: a research made in the literature; a spccificquestionnaire conceived and sent to other research projects;and lhe consideration of the end-users needs by means of the

292

application of another methodology also especialIy conceivedfor knowledge acquisition.

Taking the maio DBPMS modules into account, figure 4ilIustrates one of the graphical interfaces of the Sirnulation ofAlternatives module. The DBPMS is implemented in C++,running on a PC / Windows-NT / TCPIIP platform. .

5 MOOELLlNG THE OISTRIBUTEOBUSINESS PROCESS

For lhe system development, one of lhe importaot steps was themodeling of the Distributed Business Process (DBP). From lheDBPMS point of víew, a DBP is an object-oriented informationstructure that models lhe VE's goal, i.e. lhe final "product" inlhe Prodnet-Il ontology. It was designed in order to fulfil lheDBPMS requirements in a VE scenario. In a real case, the DBPconcept encapsulates all the suppliers, all lheorders andoperations, among other information, involved in a given VE.It differs from the traditional information that is managed by an.enterprise, which stores information (via the Prodnet's

40. SBAI - Simpósio Brasileiro de Automação Inteligente, São Paulo, SP, 08-10 de Setembro de 1999

EDIIPPC modules) about its supplier and customer orders , butanything about its "production orders" as required by the

DBPMS, i.e. about what it should do 1 produce for a givenDBP/VE.

Figure 4: Ao example of how alternative solutions are showed

The DBP structure was designed taking some concepts of theCIM-OSA project into account (AMICE , 1993). In this sense, aDBP is a hierarchica1 rnodel, composed by the basic entitiesshowed in figure 5. Regarding the usual lack of terminologystandards in the production dornain , lhe DBP modeling 1specification was strongly based on the dictionary elaboratedby the American Production & Inventory Control Society(APICS, 1983).

Figure 5: DBP model general structure

Representslhe set cI VEsInwhldl lhe VE<oordlnator Is IfMJIved m,TM 06P PcoI00jeCtIS""'--" bV0BPs.

-=tRepresents a spedtk: VElhe VE<ootdlnatochas to supervtse.ADBP tscompo:sed bV OFS.

gepeesentsa specJ(ic VE''''''''!ler (OP = Dana ln Procesf).ACPnas (oneOI'"mcre)ROs (Rec!uested Oróers).

Rcprcsents a specn: RequestedOrdet fDf a Member.AROIS<XlfTlIXl"'d bV(a ... 01)Requested llems

4 Re1Xe5ents a specltk Requested tten,A Rl ISccmpcsed bV(a sei 01)ProdU<1lon Onle<s(POs).

Deta1led desot:tlon d a PC to be supervtsed.

-? DescrlPlIon0/ ue"produc1" """"'00 10lhe PCand RI to be supervtsed .

6 CONCLUSIONSAIl the Iiterature researched and lhe conceptual work presentedin this work was actually motivated by the need to identifysome advanced coordination functionalities (ACFs) in order totry to avoid the business chaos when coordinating one or moreDBPs ofa VE.

One of lhe identified and imp1emented ACFs is the DistributedBusiness Process Management, which aims to support meansto get, provide, and to manage the information from and abouta VE, enabling the enterprises to make their logistic s moreefficiently in an integrated way. This ACF is represented by the

293

Distributed Business Process Management System (DBPMS), asystem that intends to "implernent" the emergent concepts ofSupply-Chain Management 1Integrated Logisti cs Management.

The system is modular, comprising basically three maininteractive and cooperative modules: Supervision ClausesConfiguration (to specify the rights and duties of lhe VEparticipants from the information monitorin g point of view),VE Supervisor (to monitor the DBP execution ) and DSS (aDecision Support System to help the user in the decision-making process based on lhe information obtained by the VESupervisor).

It should be noted that the DBPMS , once providing somescheduling capabilities for conflict solving, can offer somesupport to keep a "DRP" (distributed requirements planningycoherent during the execution of a given DBP and hence can beseen as a powerful module for an ERP (enterprise resourcesplanningy .

Additionally , it is important to point out that, considering: (i)the DBPMS cornplexity, (ii) the innovative aspects beinginvestigated, and (iii) the open questions still presented in theVE paradigm; lhe experiments exploited by the DBPMSmodule have their scope and scenarios Iimited and tested onlyin the Prodnet-Il (homogeneous) platforms. In terms offunctionalities, they have been oriented to the demonstrationpilot prototype (but using real data e situations), even becauseit is still not c1ear the possible overlapping between some"future" PPC functional ities and those ones the DBPMS shou1dhave.

ACKNOWLEDGEMENTSThe authors would Iike to thank CNPq (The Brazilian Councilfor Research and Scientific Development), project ProTeM-CC680120/96-3 PRODNET-ll, and European Union, Espritproject 22647 PRODNET-ll for lhe financiaI support as well asthe project partners. Special thanks to the DBPMS

Spinosa. L.M., Rabelo, a.r., KIen. A. P. Ferreira. A C.,(1998a), An oriented Decision Support System model forVirtual Enterprise Coordination, ProceedingsProlamat'98 / The Tenlh International IFIP TC5 WG-5.2 WG-5.3 Conference, Trento, Italy.

Spinosa. L.M.• Rabelo, R.J .• Klen, A P. (1998b). High-LevelCo-ordination of Business Processes in a VirtualEnterprise, in Global ization of Maniifacturing in lheDigital Communications Era of the 21st Century, Editedby Jacucci, G.• Olling, G.J., Preiss, K. and Wozny. M.•KIuwer Academic Publishers.

REFERENCES

40. SBAI- Simpósio Brasileiro de Automação Inteligente, São Paulo, SP, 08-10 de Setembro de 1(J99

development team at G-SIGMA. namely Mr. Allan Keller, Ms. industry, Proceedings Agosoft'97 / Workshop onRegina Bussolo, Mr. Rodrigo Ardigó and Mr. André Jacomino. Supply-Chain Management, Vitoria. Brazil.

Rabelo, R J.; Camarinha-Matos, L. M. (1998). Genericframework for conflict resolution in negotiation-basedagile scheduling systems, Proceedings IMS '98 - 51hIFAC Workshop on Intelligent Manufacturing Systems,Gramado - Brazil.

Afsarmanesh, R. Garita, C., Hertzberger, L.O. (1998) . VirtualEnterprises and Federated Information Sharing,Proceedings of lhe 9th IEEE Int.Conference on"Database and Expert Systems Applications",DEXA'98. Lecture Notes in Computer Science.

AMICE (1993) , CIMOSA: Open Systems Architecturefor CIM,Springer-Verlag, Berlin.

APICS (1983). The Official Dictionary of Production andInventory Management Terminology and Phrases, TheAmerican Production and Inventory Control Society,USA

Camarinha-Matos, L.M, Lima, c., Osório. AL. (1997) . TheProdnet Platform for Production Planning andManagement in Virtual Enterprises, in Proceedings oflhe 4th International Conference on ConcurrentEnterprising - ICE'97. Nottingham, UK.

Camarinha-Matos, L.M.• Afsarmanesh, H.• Garita, C; Lima, C.(1998), Towards an architecturefor virtual enterprises,Journal of Intelligent Manufacturing, Vol. 9, Issue 2, pp189-199.

Christopher, M. (1994). New directions in Logistics, Logisticsand Distribution Planning - strategies for management;edited by James Cooper, Kogan Page Limited, 2nd

edition.

Klen, A P. ; Spinosa. L.M ; RabeIo, RJ; Ferreira, A. C.(1998a). Integrated Logistics Management SupportSystein: an Advanced Co-ordination Functionality forthe Virtual Environment, to be presented in IMS'98 - 5 tlJIFAC Workshop on InteIligent Manufacturing Systems,Gramado - Brazil.

Klen. A. P.; Rabelo, R.J., Spinosa, L.M; Ferreira. A. C.(1998b), Integrated Logistics in the Virtual Enterprise:The Prodnet-Il Approach, to be presented in IMS'98 -51h IFAC Workshop on Intelligent ManufacturingSystems, Gramado - Brazil.

O'Neill, H. (1995). Decision Support in the ExtendedEnterprise, Ph.D. Thesis, Cranfield Univ .•England.

Osorio , A L.; Gibon, P.; Barata. M.(1998). Secure electroniccommerce in virtual enterprlses of SMEs. in InteIligentSystems for Mariufacturing - Multi-Agent Systems andVirtual Organizations, Eds. L. M. C àmarinha-Matos, H.Afsarrnanesh and Vladimir Marik, KIuwer AcademicPublishers, pp . 207-218.

Prodnet (1996), http://cupido.uninova.pt/-prodnet.

Rabelo, R J.; Carnarinha-Matos, L. M. (1996). Towards AgiteScheduling in Extended Enterprise, in "BaIancedAutomation Systems TI - Implementation ChaIlenges forAnthropocentric Manufacturing", Eds. L.M. Camarinha-Matos. H. Afsarmanesh, Chapman & HaIl.

Rabelo, RI.; Spinosa. L. M. (1997). A Mobile-agent-basedapproach in supply-chain supervision in the food

294