Effective Management of Dynamic and Multiple Supply Chains Ricardo J. Rabelo; Alexandra A. Pereira-Klen; Edmilson R. Klen Departments of Automation and Systems, and Mechanical Engineering Federal University of Santa Catarina, Brazil
[email protected]; {klen; erklen}@gsigma-grucon.ufsc.br Abstract: This paper presents a system called SC2 (Supply Chain Smart Co-ordination), a multi-agent decision support system which has been developed based on the business intelligence paradigm in order to better support the management of dynamic and multiple supply chains. SC2 offers an integrated environment for dealing with the production, distribution and sales chains. Via a lean interface based on XML and CORBA, SC2 can obtain reliable, timely and interoperable information from the supply chain members, comprising a number of heterogeneous information sources and legacy systems. Results are presented and discussed at the end of the paper. This work has been developed in the scope of three research projects, focusing on the highly customised supply chains. Keywords: Supply Chain Management, Multi-agent Systems, Business Intelligence. Reference to this paper should be made as follows: Rabelo, R. J. et al. (2002) ‘Effective Management of Dynamic and Multiple Supply Chains’, Int. J. Networking and Virtual Organisations, Vol. X, No. Y, pp.zzww.
1 Introduction This paper presents an approach on how a dynamic and multiple supply chains (DMSC) can be managed in a smart way, focusing on a decision support system developed to assist the enterprises in the high-level coordination of distributed business processes. With the advent of the globalisation, companies have invested a lot in ways to improve their businesses (processes) as well as to know more about their clients and suppliers in order to be more competitive. Reengineering task forces, implementation of advanced IT systems, e-commerce sites, integration of legacy systems, more powerful communication infrastructures, etc., are examples of actions that companies have taken in order to improve competitiveness. No matter the topology of the virtual organisation the companies are used (supply chain [SC], extended enterprise, virtual enterprise, etc.), all these actions have provided the companies with a basis for receiving and sending plenty of information. However, the more the companies have access to information, the more tough is the task to treat it in an efficient and smart way. In practice, the managers have been plunged into so many information that, in opposite to the core objective, it has brought even more difficulties for taking smart and agile decisions. Within this scenario, the paradigm of business intelligence (BI) comes in. In general, BI can be defined as the process to get (digital) information about the company’s entire business so that it can be used to provide the so-called competitive advantage (Malhotra, 00). Following this trend, systems to support BI have been put into the market in the past recent years, aiming to offer an integrated and global decision-making environment to the managers. These systems normally make use of information generated by other (legacy) subsystems, especially Data Warehouses, i.e. typically “passive” information. In spite of the importance of this kind of systems, they fail in the sense that they do not consider real-time data from the SCs in the course of their BI analysis. This is critical as consistent decision-making is a must nowadays, particularly in dynamic SCs, regarding that the enterprises are usually involved in several SCs simultaneously. This paper stresses the SC2 (Supply Chain Smart Co-ordination) system, a multi-agent-based decision support system that offers some contributions towards a more effective way to manage dynamic and multiple SCs. SC2 offers to the SC manager – the named SC Co-ordinator – a wider, integrated and user-friendly environment based on the BI paradigm as a support for the SC management. This work has been developed in the scope of the IST DAMASCOS [Damascos], IST MyFashion.eu [MyFashion] and IFM [Ifm] projects. The paper is organised as follows: Chapter 2 gives an overview of the difficulty to manage multiple supply chains. Chapter 3 provides a general description of the SC2 approach. Chapter 4 shows some implemented results of SC2. Chapter 5 discusses the results achieved and the next steps.
2 J. R.
Ricardo J. Rabelo, Alexandra Pereira-Klen and Edmilson R. Klen
2 Problem description One significant problem to manage SCs is to handle the enormous amount of information about and from its members (pre-suppliers, suppliers, the main producer itself, distributors, sales agents, retailers, and so on) and hence to co-ordinate the current (distributed) business they are all involved in. Sales, production status and capacities, stocks, material flow are some of the information data that should be quickly analysed to attend an even more and more demanding market. In more dynamic SCs, the core partners do not remain the same for a long period of time, and each one of the partners uses to participate in several SCs simultaneously, playing different roles sometimes. Thus, the SC management should be driven by the information related to each business, where the various relationships between the business processes can be co-ordinated as well as the material and the product flows can be constantly monitored (Filos et al., 00) (Figure 1). Figure 1
Managing multiple supply chains Production perspective
SCn
SC2 SC1 Sales perspective
Distribution perspective
However, this more advanced scenario brings much more complexity to the management model. Figure 2 illustrates how complex the global management can be even for a relatively small scenario. It is composed of ten SC members, identified by an id and name inside the main squares (1 pre-supplier, 4 suppliers, 1 principal producer, 1 sales agent and 3 retailers) and only five end-products. The arrows indicate the information flow and the material flow, relating the end-products with the respective requested orders involved in. Figure 2
Material and information flow of a supply chain
RequestedOrder 4321
Shipping Order N° 332
Requested Items:
65563 El Globo
430400560 Thread
RequestedOrder 3706
66811 FibraFio
Requested Items:
66666 Ateca
Shipping Order N°331
131313 Textile 949494 Textile
42160232
RequestedOrder 1231
Requested Items:
10 Runway 42160232 53 Cockpit 42160732 150 Chino 42160734 137 Chino 42161332 70 Chino 42161334
42161332
RequestedOrder 3893
64163 Bulgar SA
Shipping Order N° 333
Principal Producer
08088 C&A
Supplier
11101 Agent 86
39 Runway 42160232 202 Chino 42161334
Requested Items: 430300002 Button 67036011 Zipper
Pre-Supplier
03556 Roma Moda
42160734 505050 Textile 202020 Textile 313131 Textile
42161334
65131 Metallica
07999 Dandy Di
ENDPRODUCTS
42160732
63864 VictorVictor
14 Runway 42160232 22 Cockpit 42160732 125 Chino 42160734
Retailer
A Supply Chain
As it can be observed, the complexity in the management increases “exponentially” with the number of supply chain members and end-products (and the quantity of its sub-components). But this is one side, perhaps the more explicit side, of the mentioned complexity. In order to tackle this scenario, it has to be also considered that each of the involved “ERP” system (belonging to every / distributed SC actor) generates information in different interfaces, in different formats, using different terms and semantics, sometimes in heterogeous computers and operating systems, which gives rise to a huge problem of wide interoperability. Yet and still worse than this, the supporting systems for SC existing in the market let the SC manager (!)
Effective Management of Dynamic and Multiple Supply Chains
3
with the responsibility to deal will all these differences: organizing and selecting the adequate (and reliable) information; and making decision concerning a given supply chain / specific order / specific supplier. The SC2 system (Rabelo et al., 02) aims to contribute to overcome this problem, offering a global, integrated and user-friendly environment to the SC Co-ordinator, through which is possible to collect, to analyse, and to organise the information about the SCs, as well as to supervise the operational phase of the business and to support the SC Co-ordinator in decision-making.
3 The SC2 approach for co-ordination Co-operation is a key aspect for the Virtual Organisation (VO) paradigm realisation and it has been addressed in many different research projects. A co-operation involves sharing information and other resources, as well as communication, enterprise relationships and collaborative activities. Activities carried out by a company are usually organised in clusters of inter-related activities called “process” (business processes - BP), where each BP is designed to achieve an (partial) objective. When properly orchestrated, the combination of various BPs will lead to the achievement of a global VO goal. Therefore, co-operation requires co-ordination (Camarinha et al., 00). It is desirable that co-ordination can be accomplished in the most efficient way, then involving flexibility, configuration, and the usage of supporting standards as the composition of the enterprises is dynamic. There is a number of approaches to manage co-ordination in a platform for VOs. In the PRODNET-II project (Lima et al, 01), for instance, it is presented a 3 layers-hierarchical model. The highest layer indicates the activities to be carried out among two enterprises (i.e. inter Prodnet-II platforms), the middle layer specifies the activities to be done inside the Prodnet-II platform (i.e. inter-modules), and the lowest layer describes the activities to be executed inside the enterprise (i.e. by its legacy systems). In the DAMASCOS (Ferreira, et al., 00) and MyFashion projects, the approach is slightly different. The equivalent highest layer is divided into to levels (called Advanced Co-ordination Functionalities and Smart Co-ordination Functionalities), the middle layer plays the same role (called Operational Co-ordination Functionalities), and there is no functions to cover the legacy systems’ activities (it is assumed that every legacy system will put available to the Damascos’ modules the necessary data in a central database) (Figure 3). Figure 3
DAMASCOS Co-ordination Approach SC2
Smart Level
Smart
Production
Sales
Distribution
Workflow Backbone
Advanced Level
Operational Level
At the low / operational level, the co-ordination in the network is supported by a distributed workflow coordination system called Workflow Backbone (WfBB). It provides a lean way to interoperate with the DAMASCOS modules both intra DAMASCOS Suite and inter-enterprises. It is assumed that the “DAMASCOS layer” – composed of a set of modules (a suite) – is installed on the top of the every SC member, i.e. it is integrated with the enterprises’ legacy systems so that the enterprises can co-operate with each other. Besides the WfBB, the DAMASCOS suite is composed of the following modules (Figure 4): SALSA (for sales support), D3S2 (for forecasting purposes), IDLS (support for individual logistics), IPO (support for some production management) and SC2 – the module being addressed in this paper. At the advanced level, the SC2 co-ordinates the SC execution focusing on restrict/individual perspectives of parts of the SC, namely the sales, production and distribution chains. This allows the SC manager to supervise specific aspects of a given chain separately. Taking into account that a more effective SC management requires a wider view upon the SCs the enterprise is currently involved in, then a more abstract level of co-ordination is required. Thus, the smart level concentrates the most intelligent part of the SC2 system, providing a complete and integrated view of all SCs, from where the SC manager can take decisions not only based on reliable and timely information, but also based on dynamic trade-offs among the status of the different chains (Rabelo et al., 01).
4 J. R.
Ricardo J. Rabelo, Alexandra Pereira-Klen and Edmilson R. Klen
The SC2 system presents a dynamic behaviour. When the system “knows” that a given enterprise plays the role of SC Co-ordination, it allows the manager to have access to all of the systems’ functionalities (see below). By the other hand, if the enterprise is “just” a normal member, only some functionalities are enabled to be executed besides the access to limited information (previously configured). Figure 4
DAMASCOS suite & SC2 framework High-level services
Network of Enteprises
Infrastructure & Cooperation Layer (ex. DAMASCOS) SC2 WfBB
backbone ASP
SALSA
ASP
D3S2
IPO
ASP
IDLS
ERP / Legacy Systems
4 SC2 as a supporting system for Business Intelligence According to (Levi et al., 00), a number of features can characterise a system for BI support, such as: extraction and integration of data from multiples / heterogeneous sources; usage of experience and knowledge; analysis of data by multiple views; work with simulation and hypothesis; searching for causeeffects relations; transformation of passive data into useful knowledge. A complete BI system for dealing with SCs should support functionalities to comprise the SC life cycle, i.e. the activities involved in the SC creation, configuration, operation and dissolution (Camarinha-Matos et al., 00). At the current stage of developments, the SC2 system is more concentrated in the second and third phases, reflected into the following general inter-related / integrated services: ! collect, analyze, organize and put available the information about the SCs; ! offer some support in the creation and configuration phases of the supply chains; ! supervise the operation of the supply chains; ! support the SC managers in consistent decision-making. In the SC2 system these services are designed as high-level user functionalities: 4.1 Supply Chain Configurator Graphical and interactive specification of the “actors” of the SC, comprising their roles and interrelations. It is the kick-off action to activate the other functionalities. Figure 5 shows an example of its graphical interfaces, automatically created after each member of a given SC (i.e. once the SC is created) sends some predefined information. 4.2 Supply Chain Smart Map Once the SC is started, this functionality aims to offer to the SC co-ordinator a graphical and easy-to-use possibility of seeing the production, distribution and sales stages and their main characteristics. Its main graphical interface is quite similar to the one showed in the figure 5. Here it is possible to go through every member as well as to get information about the SC as a whole (including transportation). Figure 5
Example of the SC Configuration Phase
Effective Management of Dynamic and Multiple Supply Chains
5
4.3 Ad Hoc Report It aims to provide the SC co-ordinator with detailed information on specific areas of performance for consistent decision-making. There are two types of reports: - Diagnostic Report: it provides “real-time” information about a given SC so that the SC co-ordinator can feel more confident in the decision-making. It comprises the sales, the production and the distribution chains. These information are organised into several “views” / filters, such as the endproducts involved, the business processes, the SC Members per products and sub-components, sales orders, shipping orders, etc., like showed in the Figure 6. The diagnostic report’s interfaces are shown in the front of the Smart Map. In the case the co-ordinator wishes to go into the legacy systems’ database, “WWW doors” to their ASP functions are automatically instantiated (previously configured) in the interfaces. - Position Paper: it is triggered when a problem is identified during the SC operation. Alternative courses of actions (such as rescheduling) and probable consequences are drawn up based on SC Coordinator’s decisional protocols (previously modelled). Figure 6
Example of an AdHoc Report
4.4 Demand Driven
6 J. R.
Ricardo J. Rabelo, Alexandra Pereira-Klen and Edmilson R. Klen
The distribution and sales chains are an independent environment where uncertain customer demand determines independent inventory requirements. This functionality allows a faster reaction to conflicts observed in the end product’s inventory. 4.5 Distributed Business Process Management It is a decision-support functionality that helps the manager in solving problems from the production perspective (Pereira-Klen et al., 01). By means of an intensive information gathering from the SC, a periodic follow up of orders is made as well a set of alternatives schedules are suggested in the presence of problems in the SC schedule (Figure 7). Figure 7
Example of the orders Follow Up & Simulation of Alternatives
5 SC2 implementation model: a multi-agent system The SC2 is a multi-agent-based decision support system that manages the distributed business process of a dynamic supply network by means of real-time monitoring and supervision activities. It facilitates the conflict analysis and its resolution supported by (flexible and configurable) decisional protocols, providing a human-centered smart co-ordination of the SC. The whole process starts when the SC Co-ordinator wants to monitor, after a given SC is created, the execution of a particular business process (BP) at a given SC-Member’s site. It is supported by means of the sending information from each Member to the Co-ordinator (like order status, remaining process time, stock information, etc.) via the WfBB (as illustrated in Figure 4).
5.1 The SC2 Agents SC2 is composed of three classes of agents: smart supervisor, chain supervisors, and supporting agents. The first one is the agent called Smart Agent, responsible for the global SC management. The second one has three instances: Production Agent, Sales Agent and Distribution Agent, responsible for dealing with the production, sales and distribution individual chains, respectively. The third and last class is represented by four agents. The XML Agent is responsible for dealing with the communication among the agents and with the “external environment”. The Partners’ Search and Selection (PSS) Agent is responsible for finding the most suitable set of partners in the case the SC is not previously known or fixed. It is a self-contained multiagent system that combines stationary agents and mobile agents (Costa et al., 01). The SC Intelligent Configurator Agent is the agent responsible for dealing with all information required to configure dynamically and automatically a given SC and then to create the smart map (see sections 4.1 and 4.2). The Scheduler Agent represents a community of agents that are responsible to help the enterprise in the scheduling management, providing a better reactive capability to support the orders follow up (Figure 8). Figure 8
SC2 Architecture
Effective Management of Dynamic and Multiple Supply Chains
MAS-based Partners’ Search & Selection
Smart agent
SC Intelligent Configurator
SC2 WfBB ASP
SALSA
ASP
D3S 2
IPO
ASP
IDLS
Production agent Data Base
Distribution agent
DBP Model
Scheduler agents
Sales agent XML agent ORB Inter-Modules & Inter-Enterprises Interoperation
Shop floor
7
Workflow Backbone Module
...
The SC2 implementation model combines multi-agent systems, CORBA, XML, databases, and decision support systems. SC2 is a typical information-consumer system/module. Once the SC is created/configured, it receives the required information from IPO, IDLS and SALSA modules via the WfBB module (see Figure 4), regarding “contexts”. Each of these contexts corresponds to a respective XML DTD, accessed under a pooling philosophy by a SC2’s ORB specific service. Once the information is got, it is sent and managed by the XML agent. This agent parsers the XML message received and makes a matching with its internal DTD. After this pre-processing, the information is encapsulated as an object and sent to the SC2 database. The involved DBMS gets this object and stores it in a reference data model called Distributed Business Process. In parallel, the XML agent sends a message to the Production, Sales and Distribution agents so that they can be aware about the arrival of updated information. Hence the agents access the information only when they need. Along a BP operation/execution, the DBP model can be constantly read or updated, both by the XML agent and by the other SC2 agents. Results produced by the agents / SC2 are shown to the end-user via graphical interfaces. From the end-user point of view, it is totally transparent the agent being activated to provide some results.
5.2 Why Multi-agents ? The Multi-Agent System (MAS) technology has emerged as a powerful technology to support a co-operative resolution of distributed problems. In general, A MAS system corresponds to a network of problem solvers/subsystems – agents – that works together to solve problems that are beyond their individual capabilities. In practical terms, in the case of the SC2 system, the main value-added with the adoption of a MAS approach are (Rabelo et al., 01): 2 - The SC agents can be launched in distributed PCs; 2 - The integrated view of the entire supply chain is supported via a co-operative interaction among the SC agents, with flexible constraint relaxation; 2 - The SC agents can reason / negotiate about the information exchanged among them and some intelligent deductions (new knowledge) emerge from this; 2 - Other agents can be added into the SC without altering its control architecture; - Each perspective of the chain is managed in/by an autonomous way/agent.
5.3 Implementation Aspects The main strategy being applied in the SC2 development aims at having a lean, easily interoperable and standard API. For that reason, CORBA and XML technologies have been used, which means that every SC2’s agent exchange information with each other via a CORBA-XML interface. Within the SC2, the messages are modelled as KQML perfomatives, also expressed in XML. SC2 has been developed in a PC - Windows NT/2000/Xp platform, and except the PSS agent (developed in Java), the other agents have been implemented in C++. The Java agents are based on the AGLETS platform, JDK environment, and ORB ORBacus. The C++ agents are based on the MASSYVE KIT platform (Massyve), Builder C++ 5.0, and ORB TAO. The entire system is modelled in UML. The Interbase database has been utilised, with a simple object-oriented layer to provide the agents to have access to the database as objects and not as table structures.
8 J. R.
Ricardo J. Rabelo, Alexandra Pereira-Klen and Edmilson R. Klen
All software is shareware, except the C++ compiler and the operating system.
6 Conclusions, Discussions and Next Steps This paper presented the SC2 (Supply Chain Smart Co-ordination) system, a multi-agent-based decision support system that offers some support for agile decision-making for management of dynamic and multiple supply chains. This work has been developed in the scope of the IST DAMASCOS, IST MyFashion and IFM projects, whose ultimate goals is to develop open and low cost platform and management services for SMEs involved in dynamic supply chains. The functionalities showed were based on the new trends on supply chain management, on an extensive research of related projects, and on the end-users’ requirements. In the intercontinental tests carried out between the partners’ modules, the communication infrastructure, based on CORBA-Workflow-XML, achieved the main goals of being stable, flexible and open. It includes the aspect where different ORBs are working simultaneously. The SC2 system is user-friendly, highly configurable, and quite generic. The architecture has considered real constraints and data, where different strategies of interaction among the existing entities of SC2 had to be conceived, especially for efficiency purposes. Although the level of the agents intelligence is still incipient, it is foreseen a deeper development on this issue in the next future based on the feedback from the project / SC2 end-users. Yet to be soon tackled is a more dynamic way to configure the access right of information by the members of the supply chain. Other aspects important to be highlighted are related with XML / DTD models. There is a strong work to achieve an agreement among the users/systems about the information to exchange, which makes the need for reference modes a clear challenge. Every enterprise has a particular way to decide in the presence of conflicts but, sometimes, they do not know how to deal with them. About an As-is & Should-be models, a smart management of supply chain creates new requirements and the need for new models, but not even the enterprises are aware about that and how their processes should be changed.
Acknowledgements This work has been fully supported by CNPq – The Brazilian Council for Research and Scientific Development – projects DAMASCOS (n. 480101/00-0), MYFASHION (n. 68.0004/02-5) and IFM (n. 62.0051/01-9). Thanks to Mr. Fabiano Baldo, Mr. Rui Tramontim Jr., Mr. Carlos Gesser and Mr. Ricardo Schmidt for the system implementation.
References
Effective Management of Dynamic and Multiple Supply Chains 1.
2. 3. 4. 5.
6. 7. 8. 9. 10. 11. 12.
13. 14.
9
Camarinha-Matos, L.M., Afsarmanesh, H.; Rabelo, R.J., Supporting Agility in Virtual Enterprises, in E-Business and Virtual Enterprises, Eds. L. M. Camarinha-Matos, H. Afsarmanesh, and Ricardo J. Rabelo, Kluwer Academics, pp. 89-104, 2000. Costa, S.; Rabelo, R. J., Supporting the Creation of Virtual Enterprises using Mobile Agents, in Collaborative Business Ecosystems and Virtual Enterprises, Ed. L. Camarinha-Matos, Kluwer Academics, pp.371-378, 2002. Damascos, http://bart.inescn.pt/~damascos. Ferreira D., Goletz T., Ferraz, R., A Workflow-Based Approach to the Integration of Enterprise Networks, Proceedings CARS & FOF 2000, Port of Spain, Trinidad, 2000. Filos, E.; Banahan, E., Will the organisation disappear? The challenges of the new economy and future perspectives, in E-Business and Virtual Enterprises, Eds. L. M. Camarinha-Matos, H. Afsarmanesh, and Ricardo J. Rabelo, Kluwer Academics, pp. 3-20, 2000. IFM, http://www.ifm.org.br. Pereira-Klen, A.A.; Rabelo, R. J.; Ferreira, A. C., Spinosa, L.M., Managing DBPs in the Virtual Enterprise, J. Intelligent Manufacturing, V12 N2, pp. 185-197, ISSN 0956-5515, Kluwer Pub., 2001. Levi, D.S.; Kaminsky, P.; Levi, E.S., Designing and Managing the Supply Chain, McGraw-Hill, USA, 2000. Malhotra, Yogesh, Knowledge Management and Virtual Organizations, Idea Group Publishing, USA, 2000. MASSYVE, http://www.gsigma-grucon.ufsc.br/massyve. MyFashion, http://www.myfashion.org. Rabelo, R. J.; Pereira-Klen, A. A.; Klen, E., A Multi-agent System for Smart Co-ordination of Distributed Business Dynamic Supply Chains, in Collaborative Business Ecosystems and Virtual Enterprises, Ed. L. Camarinha-Matos, Kluwer Academics, pp.379-386, 2002. Rabelo, R. J.; Pereira-Klen, A.A.; Ferreira, A. C., Multi-Perspective Co-ordination of Supply Chains, Proceedings CARS & FOF 2001, pp. 484-492, Durban, South Africa, 2001. TAO, http://www.cs.wustl.edu/~schmidt/ACE.html.
