Content uploaded by Frédérique Mayer
Author content
All content in this area was uploaded by Frédérique Mayer
Content may be subject to copyright.
Universidade Federal de Santa Catarina (UFSC) – Centro Tecnológico (CTC)
Programa de Pós-Graduação em Engenharia de Produção (PPGEP) – Laboratório de Sistemas de Apoio à Decisão (LABSAD)
Campus Universitário – Trindade – Caixa Postal 476, Florianópolis – SC, BRASIL, CEP: 88040-900
Fone: ++55 (48) 3721-7065 Fax: ++55 (48) 3721-7066
http://www.ijie.ufsc.br
e-mail: ijie@deps.ufsc.br
IJIE – Iberoamerican Journal of Industrial Engineering / Revista Iberoamericana
de Engenharia Industrial / Revista Iberoamericana de Ingeniería Industrial
Periódico da área de Engenharia Industrial e áreas correlatas
Editor responsável: Nelson Casarotto Filho, Prof. Dr.
Organização responsável: Universidade Federal de Santa Catarina (UFSC)
Processo de avaliação de artigos por pares
Periodicidade: Semestral
Florianópolis, SC, vol.2, nº 1, p. 84-107 , jun. 2010
ISSN 2175-8018
Artigo recebido em 13/11/2009 e aceito para publicação em 05/06/2010
MAKING PRODUCTS ACTIVE WITH INTELLIGENT AGENTS FOR
SUPPORTING PRODUCT LIFECYCLE MANAGEMENT
Martin G. Marchetta
PhD in Engineering
CEAL, Logistics Studies and Applications Centre, School of Engineering, National
University of Cuyo, Centro Universitario, CC405 (M5500AAT) Mendoza, Argentina
mmarchetta@fing.uncu.edu.ar
Frédérique Mayer
PhD in Automatic Control
ERPI – Equipe de Recherche sur les Processus Innovatifs, ENSGSI, Institut National
Polytechnique de Lorraine, 8 rue Bastien Lepage, BP 90647 (54010) Nancy Cedex,
France
frederique.mayer@ensgsi.inpl-nancy.fr
Raymundo Forradellas
PhD in Artificial Intelligence
CEAL, Logistics Studies and Applications Centre, School of Engineering, National
University of Cuyo, Centro Universitario, CC405 (M5500AAT) Mendoza, Argentina
kike@uncu.edu.ar
ABSTRACT: Modern organization paradigms within manufacturing enterprises have arose
in last years, like Agile Manufacturing and collaboration, in order for enterprises to increase
their productivity and be more competitive in front of shorter due dates and increasing
product qualities required by customers. Most previous works on PLM and currently available
systems are usually focused on the use of additional information to support business
processes, and integrate limited information of lower-level applications (CAD, CAPP, etc).
However, little emphasis has been put on making products more intelligent during their
complete lifecycle, in order to exploit PLM information for improving their development and
management. In this paper, a framework based on intelligent agents is proposed, for giving
products active behaviors, in order to assist people involved in PLM to reduce lead times and
costs, and improving product quality. Application of the proposed framework to a product
definition example is presented as a case study.
Keywords: PLM, Active Product, Intelligent Agent, Virtual Enterprise, Concurrent
Engineering.
85
1 INTRODUCTION
Current international context has forced enterprises to optimize their operations and to
collaborate, in order to be competitive. This has yielded the need of reducing lead times and
costs, and increasing final product quality. In order to achieve these objectives, collaboration
(NOF, 2007) among peer enterprises has been more and more accepted, creating the so called
virtual enterprises (MING et al., 2008; CROXTON, 2001).
Product Lifecycle Management (PLM) is considered a key concept in order to maintain
consistency, efficiency and quality as products are created, from conception to disposal
(SAAKSVUORI; IMMONEN, 2005). PLM involves the management of product information
and the integration of business processes from birth to obsolescence of a product (SHARMA,
2005), and it is crucial for effective management of corporate intellectual capital (AMANN,
2002).
Until some years ago, integration and collaboration was implemented mostly along the
supply chain. However, currently this integration has been extended including not only
supplier and customers as partners, but also peer enterprises, as shown in Figure 1. These
partner enterprises share much information concurrently, so PLM now turns to be a
fundamental support to maintain consistency and to capitalize concurrent engineering and
collaboration benefits in these scenarios.
. . .
Direct
Supplier
Direct
Customer
Manufacturing
Enterprise
. . .
Final
Customer
Raw
Material
Supplier
Partner
Enterprise
Partner
Enterprise
Figure 1 − Virtual enterprise and supply chain integration of current manufacturing enterprises
Source: Authors
Current mixed partner-partner (BHANDARKAR; NAGI, 2000) and customer-supplier
relationships (CROXTON et al., 2001) create special issues not present before. Because
business processes now cross enterprises’ physical barriers, products information must be
shared across heterogeneous systems (with different data formats and semantics), and
86
coordination of activities needs to consider people, information with different ownership,
policies and cultures.
Previous works have focused on several aspects of PLM. However, only recently the
need of putting more intelligence on products, for making them “active” (MOREL, 2007)
during their own development and management process has been recognized, and some
efforts have started to be done. Currently, the evolution of information technologies allows
giving active behaviors to (both materialized and under development) products, which can
help companies to handle complexity, reducing inconsistencies and optimizing product’s
definition and manufacturing activities.
In this paper, a framework for making products actively involved in their own
development and management within PLM is proposed, which puts emphasis on an
applications architecture based on intelligent agents. The framework is aimed at defining the
relevant entities, interaction types and information sources for proactive support to PLM, and
explores suitable techniques and technologies to implement it.
2 PREVIOUS WORKS ON INFORMATION EXPLOITATION WITHIN PLM
PLM has evolved in recent years along different research directions, focused on the
creation, storage and use of product information. Among these aspects, the most relevant one
from the point of view of this work is the product information exploitation along its lifecycle.
This aspect raises some issues, such as interoperability due to different information formats
and semantics, and the difficult to guarantee completeness and consistency (SAAKSVUORI;
IMMONEN, 2005).
Some works propose limited support and coordination mechanisms both within a single
activity and between activities and partner enterprises. Assistance on specific activities [e.g.
CAPP (TONG; LI; YUAN, 2008)] within a single company has been proposed. Support for
collaboration between customers, suppliers and partner enterprises focused on specific
activities has also been proposed (MING et al., 2008).
Another application for information exploitation is related to change propagation.
Consistency maintenance of both product information (MA; CHEN; THIMM, 2008) when
changes are made through product development) and design documents (SHIAU; WEE,
2008) have been addressed.
Researches related to intelligent products have also been developed, on which the idea
is connecting the physical products with their counterparts within information systems, based
on intelligent agents, holons, RFID and other technologies (MEYER; FRÄMLING;
87
HOLMSTRÖM, 2009; VALCKENAERS , 2009; YANG; MOORE; CHONG, 2009). The use
of different kinds of agent technologies (holonic approaches, multi-agent systems, etc)
(MARIK; LAZANSKY, 2007), composition of services through semantic interoperability
(CONTRERAS; SHEREMETOV, 2008) and multi-agent systems for planning and
coordination (FORGETA; D’AMOURSA; FRAYRET, 2008) are examples of recent efforts
for putting more automated support to PLM.
The kind of applications taking advantage of PLM information is restricted to a few
activities and to specific problems. Putting intelligence on products is starting to be explored,
but there is no general framework defining the overall structure of interactions, involved
entities and information sources for exploiting information within PLM without specifying a
specific aspect or activity to be supported.
3 BUSINESS PROCESS MODEL FOR PLM AND SCM
In previous works, the product role within PLM has been to be an information hub,
concentrating all data required by different activities along the PLM. This limited view allows
people from companies to get answers to questions regarding product information, but only
people take the initiative for exploiting this information. PLM business processes have also
been studied, but there has been a separation between these two dimensions (PLM business
processes and information to support them), what reduces the benefits obtained from the
implementation of PLM.
Current technologies allow putting active behaviors inside information systems, what
turns them into an active actor during PLM activities. This gives the possibility of having
additional benefits as product information is generated and stored, like identifying
optimization opportunities or detecting potential risks, rather than only passively answer
questions when people need it.
Several processes must be carried out by an industrial enterprise having relationships
both along the supply chain (with suppliers and customers), and with partners during product
development and/or production. Considering this scenario, Figure 2 depicts a logistics and
production oriented business process model.
88
Figure 1 − Overview of Business Process Model for joint PLM and SCM
Source: Authors
89
The model is based on two different (and somehow complementary) views of business
processes in logistics (CROXTON et al., 2001; FRAZELLE, 2002). Processes of a product
development and manufacturing company are shown inside the rectangle. The supplier, the
customer and the partner enterprises are external actors. Partners are enterprises which design
other product’s components, or that manufacture products designed by the company.
Two types of customer orders may be received by the enterprise: new product
development orders and orders for existing products. In addition to this, at any moment the
customer may ask for information about its order’s state and/or the state of the development
process of the new product. All these information flows are handled by the Customer Service
Management process.
The Order Delivery process takes care of coordinating all activities to put products in
the location accorded with the customer. The Warehousing process includes all warehouse
management activities, such as picking, put away, storage, etc.
Individualized Product Definition process groups all new product development (NPD)
activities, including turning customer needs into requirement specifications, designing the
product, creating manufacturing specifications, etc. This process is at the core of our model,
since it handles consistency and coordination both inside the company (with other business
processes) and with the environment (partners, suppliers and customers).
In this work, the system for supporting coordination and consistency maintenance
within this process is modeled as an active one since it uses all the available information to
proactively help managing the complete product’s lifecycle.
The Production Planning process aims at accommodating demand to manufacturing
resources considering firm orders, forecasts and manufacturing process plans indicating
routings, bills of materials, time and other resources. Demand/Inventory Management
forecasts demand in order to support procurement and production planning. Procurement is
concerned with handling of buying orders. Finally, the Manufacturing process is in charge of
products creation using manufacturing resources. It mainly uses product specifications from
Individualized Product Definition and the amount of units to be produced, when they should
be produced and with which resources from Production Planning.
Integration with partner enterprises is mostly carried out by means of the Individualized
Product Definition process, yielding a great number of interactions of all types, including
interactions along the supply chain, interactions with partner enterprises, and also within
several activities inside the process itself.
90
Figure 2 − Individualized Product Definition’s functional diagram
Source: Authors
91
Figure 3 presents the Individualized Product Definition functional diagram. The process
is composed of 5 main activities: Product Specification Definition, Conceptual Design,
Product Detailed Design, Product Engineering and Process Planning. In a concurrent
engineering environment, these activities are carried out in an overlapped way, rather than
sequentially. Thus, there is a complex (and usually asynchronous) information exchange
among them.
3.1 Hierarchy of product-automated support along product lifecycle
Integrated PLM within a concurrent engineering/virtual enterprise paradigm involves
interactions among several types of entities and at different abstraction levels, such as inter-
activity interactions within a single company (MA et al, 2008) and inter-company vs. intra-
company scenarios (SHIAU; WEE, 2008). Since each interaction type entails different issues
(e.g. information ownership, concurrency, etc), and also presents different potential
opportunities for realizing active-product’s benefits, a structured categorization of these
interactions is needed.
Both the “business processes” and the “virtual enterprise” dimensions are considered
here for the proposed classification. Another aspect to be considered is that the product
development process, as defined in this article, includes several internal activities, and also
has relations with other business processes within a single organization. Additionally,
individual activities themselves present opportunities for improving the development process
by means of an active product approach. As a result, a 4-level interactions hierarchy, depicted
in Figure 4, is proposed here as follows:
1. Inter-process/multiple company interactions: it is represented in the figure by thick
continuous arrows. It involves information exchanges between processes in different
companies within the virtual enterprise. Information ownership must be taken into
account, since interactions cross a single enterprise barriers.
2. Inter-process/single company interactions: this case is depicted by gray arrows in the
figure. It is related to interactions among business processes within a single company.
3. Inter-activity interactions: inside a single business process, there exists the need of
coordinating activities which contribute to its value chain. This is especially true when
activities whose results impact on the other ones are executed concurrently. This
interaction type is shown in the figure with dotted arrows, and activities are
represented by ellipses.
92
4. Intra-activity interactions: Within a single activity the number of people involved may
be very small, and so the information exchanged, so the most important thing an
active product may contribute with is exploitation of information in order to optimize
results of that activity.
Virtual Enterprise
Partner
Company
. . .
. . .
Business
Process
i
Business
Process’
1
Business
Process
2
Business
Process
n
A
1
A
2
Business
Process
m
Business
Process
1
Inter-process/
single company
interaction
Inter-activity
interaction
Intra-activity
interaction
Inter-process/
multiple company
interaction
Figure 4 − Hierarchy of active-product’s interventions along its lifecycle
Source: Authors
These interactions are points for potential contribution of automated support by an
active product. An active product is similar to an expert advisor who actively integrates all the
information across the product’s lifecycle, resulting in a PLM system of PLM inter and intra
systems, from business to manufacturing and across all the partners, as a whole (MAIER,
1998). For inter-process situations (types 1 and 2), most important benefits include change
impact analysis, propagation/notification of changes to (and only to) relevant people, global
optimization, risk assessment, feedback on project evolution, etc.
Considering inter-activity interactions (type 3), similar benefits can be obtained, but
restricted to activities within a single business process. Finally, inside a single activity (type
4), local multi-objective optimization can be supported by the product, as well as other aids
such as know-how acquisition through machine learning (MARCHETTA; FORRADELLAS,
2006a, MARCHETTA; FORRADELLAS, 2006b), plan recognition for identifying user’s
intentions, retrieval of patterns used in the past in a mixed-initiative approach (ALLEN;
GUINN; HORVTZ, 1999), automated task completion (MARCHETTA; FORRADELLAS,
2007), etc.
93
3.2 Applications architecture
Software applications that support product development have emerged independently
from the need for automated support of individual activities (CAD, CAM, scheduling, CRM,
etc), which has generated issues related to interoperability, data formats, lack of coordination
and collaboration support, etc. Figure 5 shows a typical applications architecture used within
product development companies within virtual enterprises.
94
Figure 3 − IT applications architecture typically used within virtual enterprises for product development
Source: Research
95
Because of the great diversity in the nature of activities covered along the PLM, it is
difficult to create a single software application for supporting all of them. Besides, many
companies have made great investments in solutions to individual problems. Moreover,
within virtual enterprises it is likely that different software solutions be used to solve the same
problems (e.g. different CAD/CAPP/CAM systems). Because of these reasons,
interoperability between existing applications has turned to be a recurrent issue.
Usually PLM systems provide interoperability by being an information backbone
(DENKENA et al., 2007). Standards are very important for achieving interoperability
(RACHURI et al., 2008), and those related to information exchange (e.g. ISO 10303), have
supported this trend.
The problem with this approach is that much information must be converted to other
formats in order to pass it from one system to another, which usually produces some semantic
information loss. Besides, these information exchanges are usually made by means of manual
coordination mechanisms, and through external support applications such as e-mail. Product
information should be handled within the systems as much as possible, avoiding conversions
and re-conversions, and non-structured communication mechanisms.
Especially because of the common current “pull strategy” of product information
exploitation, many optimization opportunities as well as early detection of global problems
can be missed. These improvement opportunities may be related to any of the interaction
levels mentioned in section 3.1. For example, by enriching product information with the
rationale for its design, its manufacturing plan, etc. automated reasoning and optimization
techniques may be used in order to improve its quality and profitability. Techniques like
artificial intelligence planning and constraint satisfaction may be used to produce alternative
solutions to design, manufacturing planning or production scheduling problems, and other
techniques like simulated annealing, may be used for optimizing solutions. Patterns across
different product development activities may be captured through machine learning (e.g.
through learning of decision trees, association rules, etc) and be later used.
Figure 6 depicts the proposed global IT architecture. For simplicity, only entities
relevant to the individualized product definition process are included in the figure. The
architecture aims at providing a structured integration infrastructure to: (1) Support
development of applications with proactive behaviors; (2) Support enhanced interoperability
among heterogeneous systems (different applications, from different vendors, owned by
different organizations, etc); (3) Reduce information exchanged in non-structured formats.
96
Reduce information re-work, by including rationale and semantics behind product
development decisions within the PLM system.
97
Figure 4 − Proposed IT applications architecture for supporting active-products
Source: Research
98
This architecture can be accommodated to be independent of the particular software
solution or application vendor. In order to realize the proposed objectives, as much semantic
information as possible must be put into the system. This will allow not only pointing to
product design and manufacturing specification files, but also to automatically reason on
them, make improvement suggestions, identify impact of distributed changes made along
different activities, and support other kinds of improvements such as global optimization.
It is common nowadays to have interfaces to access information stored within systems
through web services and Service Oriented Architectures (SOA) standards. This allows
reusing knowledge already available in current applications, such as CAD, CAPP, CAM,
ERP, etc, and gives the architecture the capability of taking advantage of information
distributed across different locations. As depicted in Figure 6, coordination, collaboration and
information exchanges are possible between organizations through interfaces exposed as
services.
Since there is no system able to support all relevant product information, global
optimization can be achieved by means of collaborative optimization techniques. In the
proposed framework, the main idea is to consider the product at the core of its lifecycle
management, not only as a reactive entity, but also as a proactive one capable of identifying
opportunities to be exploited and to suggest how to do that.
In figure 6, an entity called Product Agent is at the center of the architecture, and plays
the role of an active product. Product Agent is an Intelligent Agent (RUSSEL; NORVIG,
2002; Wooldridge; Jennings, 2002), whose environment is composed of both reactive (e.g.
other applications) and proactive entities (e.g. human users, teams or other artificial agents)
involved in PLM activities and processes. The Product Agent acts as an automated expert
connected to all the applications supporting PLM activities, which is capable of identifying
events that take place on the environment and to act as a consequence. It has also the ability to
communicate with other agents when not enough information is available to make (or
suggest) decisions, such as when information from different partners must be put together.
This also allows managing information property issues, since there is no single agent having
all the information but a set of agents, each one having access to a part of it, which
communicate with each other to exchange data.
Thus, the system moves from isolated automation islands towards a PLM integrated
system whose organization can be compared to the notion of system of systems (MAIER,
1998). This concept is related to adaptable distributed systems which interact with each other,
99
resulting in productivity and functionality that are greater than that provided by the sum of
individual systems. In the proposed framework, the notion of system of systems is considered
as a potential artifact (MAYER; AUZELLE, 2007) in order to design a PLM system having
active behaviors, supported by the underlying systems, data bases, information models and
processes.
Several issues must be solved in order to achieve the enhanced capabilities mentioned
above. First, this new system must be able to automatically exploit information created and
processed by people scattered across the virtual enterprise. Second, the system must be
capable of communicating with other similar systems, which may represent other’s interests
and intentions. Third, it must also be able to access information from both autonomous and
non-autonomous systems, such as CAD and CAPP systems. And finally, it should do all of
this autonomously and dynamically, which means that it must detect special situations and act
accordingly without having continuous and direct human intervention. Combined to our
proposed modeling framework, intelligent agents are a suitable technology for supporting
this.
3.3 Proactivity and intelligent agents
The implementation of a PLM framework, like the one described in previous sections,
requires technology having some properties that give it the ability to provide enhanced PLM
information exploitation. The active-product exchanges information both in response to user
demands (regular queries like current development stage, estimated release dates,
optimization recommendations etc.), and proactively without human intervention (e.g.
proposing optimizations, etc).
In order to proactively assist teams and organizations in PLM, a product needs to have
the ability to detect relevant events and to make decisions automatically. Wooldridge and
Jennings (WOOLDRIDGE, 1995) discussed several definitions of intelligent agents including
a “weak” one in term of 4 properties an intelligent agent should have: autonomy, social
ability, reactivity and proactivity.
Autonomy is related to the level to which the agent can act by itself without human
intervention. Since one of the most complex tasks within concurrent engineering in large
projects is to detect relevant events (such as collateral effects of changes or optimization
opportunities) and to coordinate activities, an intelligent product agent should be as
autonomous as possible.
100
Social ability allows an agent to communicate with other agents and also with humans.
This is useful for improving interoperability since partner enterprises do not need to use the
same applications or data formats, as long as the agents speak the same language (both
semantically and syntactically).
Reactivity means that agents should receive requirements from human users, and
respond to them. On the other hand, proactivity is related to goal-directed behaviors, which
means that the product agent may take the initiative when it has detected events in the
environment.
As an example of the role of these properties during PLM, consider the case of two
partner enterprises that develop 2 components of the same product which are related through
assembly interfaces. One of these components is being subject to product engineering, and a
structural problem is found during this activity. Therefore, the engineering team suggests a
design modification of the component. This design suggestion is a relevant event that must be
detected by the product agent, since it requires change propagation and coordination of
activities. Thus, when this design change is suggested, the product agent can search its
knowledge base for finding alternative design patterns associated with the corresponding
product requirement, and suggests them to the design team. When the design team makes a
decision (which is another relevant event), changes made on the component are analyzed by
the product agent and the required changes are propagated to the involved people, both inside
the organization (engineering and manufacturing planning teams), and in partner
organizations.
4 CASE STUDY
In order to illustrate how the product-agent acts within the proposed framework, some
situations within a simple case study are commented here, on which potential techniques for
implementing automated assistance are explored. Figure 7 shows a flowchart diagram of
activities required for a product development project, starting with the detailed design (for
simplicity previous activities are not shown in the diagram). In the figure, development teams
within an organization interact with partner organizations and also with the product agent.
101
Figure 5 − Flowchart diagram for the Individualized Product Definition supported by Product Agent
Source: Research
102
After a product requirement specification is available, the first stage involves the
creation of a preliminary design. As soon as a preliminary design is created, most of other
activities can be started. Product engineering and manufacturing engineering team’s activities
can be performed in parallel. Changes in product’s design or manufacturing specifications are
relevant events that trigger the product’s agent analysis mechanisms. From these analyses,
assistances like automated change notification and propagation to the affected people are
carried out. These changes may be directly introduced by users, or these users may accept
improvement suggestions proposed by the product agent. Moreover, changes may be
produced in processes within a single company, or in teams of a partner. A suitable Product
Information Model (PIM) is necessary in order to determine which people and information
elements are affected by some change. Product features (design features, manufacturing
features, etc) are a useful tool to relate different product views (e.g. which design features
respond to product’s requirement, which manufacturing features are related to a design feature
and which product’s geometric elements are involved). This helps to determine the impact
that decisions made in some activity have on other activities. Knowledge based techniques
(e.g. expert systems, constraint satisfaction, etc.) may help to determine change impacts and
support coordination.
Global optimization may involve improvements on several aspects of a product, such as
design or manufacturing plan. Global product optimization involves collaborative
optimization or information sharing, since it requires the consideration of product data from
several processes and partners. The more information on resources, patterns and good
practices is available, the better optimizations may be synthesized and suggested by the agent.
For example, if the design rationale has been formalized and stored in the PIM,
automated reasoning (through ai-planning, expert systems, constraint satisfaction, etc) may be
used to suggest alternative design patterns fulfilling the same product’s requirements. These
patterns may be hand-coded within the knowledge base, or may have been learned from past
experiences. One way to formalize and store the design rationale is to associate design
patterns (like design parametric features) with their functionality, and to associate also
requirements with functionalities that may fulfill them, by representing this knowledge in
some logic language, suitable for performing automated reasoning. Global optimization may
allow the product agent to consider the design rationale of components beyond a single
partner, by proposing improvements crossing enterprise’s barriers.
Local assistances involve making easier some tasks such as feature identification and
process planning, and suggesting alternatives for improving results (e.g. design and process
103
planning optimizations). For example, an agent for the CAPP activity has been developed in
previous works (MARCHETTA; FORRADELLAS, 2006a; MARCHETTA
FORRADELLAS, 2006b; MARCHETTA; FORRADELLAS, 2007). This agent is capable of
learning process planning patterns by observing decisions made by the manufacturing
engineer. It also has the ability of synthesizing new process plans based on the available
manufacturing resources and the product’s design, as well as identifying manufacturing
features from the product’s geometry. While the manufacturing engineer is working on a new
process plan, the agent tries to identify its intentions using plan recognition techniques and
then the agent suggests different ways of completing the plan, including using the same
known patterns (learned from past experiences) and creating new process plans fulfilling the
same goals but optimizing some criterion (time, cost, etc). Figure 8 shows an overall schema
of this agent.
After the final design has been obtained, final engineering analysis and process planning
performed for completing the product specification. This information will later be used for
production scheduling and manufacturing in order to create the physical product.
Figure 8 − Structure of a mixed-initiative system based on an intelligent agent for CAPP
Source: Research
5 CONCLUSIONS AND FUTURE WORK
In this paper a framework including a business process model, an architecture of
applications and the use of modern technologies, particularly intelligent agents, was proposed
to improve and enhance information exploitation along PLM. The presented framework aims
at providing a suitable model to expand the participation of the product along its own
lifecycle, in order to reduce time-to-market, improve final quality and reduce coordination
and consistency issues in virtual enterprises embedded in the supply chain.
104
One contribution of this work is the integrated treatment of both the partnership within a
virtual enterprise setting seen from the product development stages, and the Supply Chain
Management integration, within the Business Process Model proposed.
Another contribution is the proposal of architecture of applications to support PLM by
means of an intelligent product agent, having the properties of autonomy, social ability,
reactivity and proactivity.
In addition to that, a hierarchy of interactions including the different kinds of assistance
an active product agent may give to product development teams is also proposed, which is
exemplified with some concrete coordination, local and global optimization opportunities.
This allows generalizing the joint enterprise and application architecture to apply these and
other automated assistances, instead of concentrating on a single one (such as change
propagation).
Finally, this framework constitutes a starting point for implementing an integrated
architecture for PLM, including suggestions on the use of modern technologies such as
intelligent agents and other artificial intelligence and optimization techniques.
From the enterprise and application architecture point of view, future work will include
a refinement on the business process model introduced here, as well as a product information
model, which is required to support a product-agent within PLM. Additionally, application of
the same active-product concept to previous and later stages, along manufacturing and
logistics administration may be explored, to give the product the ability to coordinate its own
production in the shop floor, storage, transportation, raw materials procurement, etc. In this
approach a product would have an immaterial existence (the product agent, including its own
information and knowledge), and a material one (the product that is physically constructed,
stored, transported, etc). This enables new production management schemas to explore, such
as distributed manufacturing and logistics coordination.
From the framework realization point of view, some research has already been done in
the local optimization and assistance (the fourth level of the proposed interactions hierarchy),
mostly in the CAPP activity (MARCHETTA; FORRADELLAS, 2006a; MARCHETTA;
FORRADELLAS, 2006b; MARCHETTA; FORRADELLAS, 2007). Further work is needed
in order to implement assistance along the complete proposed hierarchy. In the near future,
the implementation of the CAPP local assistance agent will be finished, and research will
continue for realizing agent support to other activities, especially on coordination, global
optimization and change propagation, following the guidelines defined in this paper.
105
ACKNOWLEDGEMENTS
This research was jointly developed by the School of Engineering (UNCuyo, Argentina)
and ENSGSI (INPL, France), and partially financed by the PREMER F-599, ARFITEC ARF-
08-04 programs, and CONICET PhD fellowship.
REFERENCES
ALLEN, J. E.; GUINN, C. I.; HORVTZ, E. Mixed-Initiative Interaction. IEEE Intelligent
Systems and Their Applications, v. 14 n. 5, p. 14-23, 1999.
AMANN, K. Product lifecycle management: empowering the future of business, CIM Data,
Inc.; 2002.
BHANDARKAR, M. P.; NAGI, R. STEP-based feature extraction from STEP geometry for
Agile Manufacturing. Computers in Industry, v. 41, p. 3-24, 2000.
CONTRERAS M. SHEREMETOV L. Industrial application integration using the unification
approach to agent-enabled semantic SOA. Robotics and Computer-Integrated
Manufacturing, v. 24, p. 680-695, 2008.
CROXTON, K. L.; García-DASTUGUE, S. J.; Lambert, D.; ROGERS, D. The Supply Chain
Management Processes. The International Journal of Logistics Management, v. 12, n. 2, p.
13-36, 2001.
DENKENA, B.; SHPITAINI M.; KOWALSKI, M. P.; MOLCHO, G.; ZIPORI, Y.
Knowledge Management in Process Planning. CIRP Annals – Manufacturing Technology v.
56, n. 1, p. 175-180, 2007.
FORGETA P., D’AMOURSA S.; FRAYRET J. Multi-behavior agent model for planning in
supply chains: An application to the lumber industry. Robotics and Computer-Integrated
Manufacturing, v. 24, p. 664-679, 2008.
FRAZELLE, E Supply Chain Strategy. McGraw Hill, 2002.
MA, Y.; CHEN, G.; THIMM, G. Change propagation algorithm in a unified feature modeling
scheme. Computers in Industry, v. 59, p. 110-118, 2008.
MAIER, M. W. Architecting principles for systems-of-system. Systems Engineering, v. 1, n.
4, p. 267-284, 1998.
MARCHETTA M.; FORRADELLAS, R. Artificial Intelligence Planning for Generative
Computer Aided Process Planning. 19º International Conference on Production Research,
Valparaiso, Chile, 2007.
MARCHETTA, M.; FORRADELLAS, R. A new model for automatic generation of plan
libraries for plan recognition. Brazilian Journal of Operations and Production
Management, v. 3, n. 2, p. 5-19, 2006a.
MARCHETTA, M.; FORRADELLAS, R. Supporting Interleaved Plans in Learning
Hierarchical Plan Libraries for Plan Recognition. Inteligencia Artificial: Revista
106
Iberoamericana de IA, (AEPIA, Asociación Española para la Inteligencia Artificial), n. 32, p.
47-56, 2006b.
MARIK V.; LAZANSKY J. Industrial applications of agent technologies. Control
Engineering Practice v. 15, p. 1364-1380, 2007.
MAYER, F.;AUZELLE, J. P. Is system of systems a candidate rationale artifact for entreprise
information-intensive system modeling?, 9th MITIP International Conference on The
Modern Information Technology in the Innovation Processes of the Industrial
Enterprise, Florence, Italie, 2007.
MEYER G.; FRÄMLING K.; HOLMSTRÖM J. Intelligent products: a survey, Computers
in Industry, v. 60, p. 137-148, 2009.
MING, X. G.; YAN, J. Q.; WANG, X. H.; LI, S. N.; LU, W. F.; PENG, Q. J.; MA, Y. S.
Collaborative process planning and manufacturing in product lifecycle management,
Computers in Industry 59 (2008) 154-166.
MOREL, G.; VALCKENAERS, P.; Faure, J. M.; PEREIRA, C. E.; DIEDRICH, C.
Manufacturing plant control challenges and issues. Control Engineering Practice, v. 15 n. 11,
p. 1321-1331, 2007.
NOF S.Y., Collaborative Control Theory for e-Work, e-Production, and e-Service. Annual
Reviews in Control, v. 31, n. 2, p. 281-292, 2007.
RACHURI, S.; Subrahmanian, E.; BOURAS, A.; FENVES, S. J.; FOUFOU, S.; SRIRAM, R.
D. Information sharing and exchange in the context of product lifecycle management:
Role of Standards, Computer-Aided Design 40 p. 789-800, 2008.
SAAKSVUORI, A.; IMMONEN, A. Product lifecycle management, 2nd edition, Springer
Berlin, 2005.
SHARMA, Collaborative product innovation: integrating elements, of CPI via PLM
framework. Computer-Aided Design, v. 37, p. 1425–1434, 2005.
SHIAU, J.; WEE, H. M. A distributed control workflow for collaborative design network.
Computers in Industry, v. 59, p. 119-127, 2008.
TONG, Y.; LI, D.; YUAN, M. Product lifecycle oriented digitization agile process
preparation. Computers in Industry, v. 59, 145-153, 2008.
VALCKENAERS P., SAINT GERMAIN B., VERSTRAETE P., VAN BELLE J., HADELI;
VAN BRUSSEL H. Intelligent products: Agere versus Essere. Computers in Industry, v.
60, p. 217-228, 2009.
WOOLDRIDGE, M. An Introduction to MultiAgent Systems. John Wiley & Sons Ltd,
2002.
WOOLDRIDGE, M.; JENNINGS, N. R. Intelligent Agents: Theory and Practice. The
Knowledge Engineering Review, v. 10, n. 2, p. 115-152, 1995.
107
YANG X.; MOORE P.; CHONG S. K. Intelligent products: From lifecycle data acquisition to
enabling product-related Services. Computers in Industry, v. 60, 184-194, 2009.
COMPLEMENTARY BIBLIOGRAFHY
ISO 10303-1: Industrial automation systems and integration – Product data representation
and exchange, Part 1: Overview and fundamental principles, International Standards
Organization, 1994.
RUSSELL, S.; NORVIG, P. Artificial Intelligence: A Modern Approach, 2nd Ed., Prentice
Hall, 2002.
SCHUH, G.; ROZENFELD, H.; ASSMUS, D.; ZANCUL, E. Process oriented framework to
support PLM implementation. Computers in Industry, v. 59, p. 210-218, 2008.
TRAPPEY A. J. C., LU T.; FU L. Development of an intelligent agent system for
collaborative mold production with RFID technology. Robotics and Computer-Integrated
Manufacturing, v. 25 (2009) p. 42-56, 2009.
