International journal of computer integrated manufacturing , tập 23, số 2, 2010

International Journal of Computer Integrated Manufacturing
Vol. 23, No. 2, February 2010, 101–112

Agent-based workflow management for RFID-enabled real-time reconfigurable manufacturing
YingFeng Zhanga,b, George Q. Huanga*, Ting Qua and Oscar Hoa
a

Department of Industrial and Manufacturing Systems Engineering, The University of Hong Kong; bThe State Key Laboratory for
Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an, China
(Received 22 April 2009; final version received 27 October 2009)
Recent developments in wireless technologies have created opportunities for developing reconfigurable wireless
manufacturing systems with real-time traceability, visibility and interoperability in shop-floor planning, execution
and control. This paper proposes to use agent-based workflow management as a mechanism to facilitate interactions
among RFID-enabled reconfigurable manufacturing resources. A production process is modelled as a workflow
network. Its nodes correspond to the work (process), and its edges to flows of control and data. Nodes are
represented as agents and edges as messages. As a sandwich layer, agents wrap manufacturing services around a
work-cell and their operational logics/intelligence for cost-effectively collecting and processing real-time
manufacturing data, forming so-called work-cell gateways. A reference framework for a shop-floor gateway is
proposed based on the three key components: Workflow management, manufacturing services universal description,
discovery and integration (namely MS-UDDI) and work-cell agents. Work-cell agents are packaged, registered and
published at MS-UDDI as web services which are easily reused and reconfigured in the workflow for a specific
production process. Finally, a prototype system is presented to demonstrate how the proposed method is used to
define and execute a real-time reconfigurable manufacturing project.
Keywords: workflow management; multiple agent systems; reconfigurable manufacturing; real-time wireless
manufacturing; RFID/auto ID

1. Introduction
With the increasing competition in the global marketplace, manufacturing enterprises have to strive to
become responsive to business changes which have
further impacts upon production goals and performance at the shop-floor level. Many business problems

manufacturing enterprises are facing now are caused by
lack of timely, accurate, and consistent shop-floor
manufacturing data. The infrastructure and tools need
to be designed and developed for reconfiguring a
manufacturing process to visualise, monitor and control
its real-time execution according to various production
orders. Therefore, it is essential for manufacturers to
upgrade their capabilities with advanced manufacturing
technologies (AMT), in terms of both software and
hardware technologies. Recent developments in wireless
sensors, communication and information network
technologies (e.g. radio frequency identification –
RFID or Auto-ID, Bluetooth, Wi-Fi, GSM, and
infrared) have nurtured the emergence of reconfigurable
wireless manufacturing (WM) (Huang et al. 2009) as
next-generation manufacturing systems (NGMS).
The concept of RMS (reconfigurable manufacturing systems) was introduced in the mid-nineties by
Koren et al. (1999) as a cost-effective response to

*Corresponding author. Email:
ISSN 0951-192X print/ISSN 1362-3052 online
Ó 2010 Taylor & Francis
DOI: 10.1080/09511920903440354


market demands for responsiveness and customisation.
It was pointed out that RMS was designed at the
outset for rapid change in structure, as well as in
hardware and software components, in order to
quickly adjust production capacity and functionality

within a part family in response to sudden changes in
market or in regulatory requirements. The ultimate
goal of RMS is to utilise a systems approach in the
design of manufacturing process that allows simultaneous reconfiguration of (1) the entire system, (2) the
machine hardware and (3) the control software.
Radunovic (1999) proposed an innovative RMS
paradigm to dissolve the hard borders between hardware and software and join the potentials of both.
Zhao et al. (2000) considered a RMS as a manufacturing system in which a variety of products required by
customers are classified into families, each of which is a
set of similar products, and which correspond to one
configuration of the RMS. Mehrabi (2000) proposes
five key characteristics for RMSs. They are modularity, integrity, convertibility, diagnosability and customisation. Yigit and Usloy (2002) describe a modular
structure to accommodate new and unpredictable
changes in the product design and processing needs
through easily upgrading hardware and software


102

Y. Zhang et al.

rather than the replacements of manufacturing system
elements such as machines.
Despite of significant progress achieved in the
above researches in utilising RMS, a breakthrough is
yet to come in reality. One critical hurdle is the lack of
a RMS infrastructure imminently required for manufacturing enterprises to achieve real-time and seamless
dual-way connectivity and interoperability between
application systems at enterprise, shop-floor, work-cell
and device levels. The following questions are open for

investigation:
(1) How rapidly and flexibly to define reconfigurable manufacturing resources (which tasks/
processes should be allocated to which workcells) according to the real-time manufacturing
data when production orders are changed, and
to monitor the real-time manufacturing progress during the execution stage?
(2) How to wrap applications and services around
work-cells and their flexible manufacturing
objects so that they can be easily reused and
reconfigured for different customer orders that
require different production processes?
(3) How to capture the real-time manufacturing
data by installing Auto-ID devices as manufacturing services into manufacturing devices?
This research adopts and develops three important
concepts in order to address the above questions in
building up an RMS infrastructure. They are workflow
management, multiple agent system, and automatic
identification and data capturing technologies. Workflow management (Elmagarmid and Du 1998) is a wellknown technology supporting the reengineering of
business and information processes, involving two
main stages: definition and execution. The concept can
be readily extended to define and execute manufacturing
processes to implement reconfigurable manufacturing.
Agent technologies provide necessary autonomy, flexibility and reconfigurability (Sikora and Shaw 1998,
Macchiaroli and Riemma 2002) in reconfigurable
manufacturing. Agents are used in this research to
wrap the work-cell applications related to manufacturing objects such that they can be easily reused and
reconfigured in a manufacturing process defined as a
workflow system. Auto-ID technologies, such as RFID,
can be used to capture the real-time manufacturing data
by deploying RFID devices (Readers and Tags) to
manufacturing resources (Huang et al. 2007). This

RFID-enabled real-time RMS provides a new paradigm
for production systems which accommodates higher
flexibility in terms of product volumes and types.
This paper discusses the challenges of integrating
the above three concepts into a coherent infrastructure

framework for implementing real-time RMS. The rest
of the paper is organised as follows. Section 2 reviews
the literature of workflow management, agent-based
manufacturing and automatic data capturing for
wireless manufacturing. Section 3 outlines a systematic
overview of the real-time reconfigurable manufacturing infrastructure. A referenced framework of shopfloor gateway is proposed in Section 4. Section 5
presents the concept of agent-based workflow management and its key enabling technologies. A case for
shop-floor assembly line is designed and demonstrated
in Section 6. Conclusions are drawn in Section 7.
2.

Literature review

Three streams of literature are relevant to this research.
They are workflow management, agent-based manufacturing applications and automatic data capturing
for real-time manufacturing.
2.1. Workflow management
Workflow management is a diverse and rich technology and is now being applied over an ever increasing
number of industries. Hollingsworth (1994) defines
that a workflow process is a coordinated (parallel and/
or serial) set of process activities that are connected in
order to achieve a common business goal. According
to Schal (1996), the workflow management system is
used to define, manage, and perform ‘workflows’

through the execution of software, whose order of
execution is driven by a computer representation of the
workflow logic. Workflow technology is increasingly
used to manage complex processes in e-commerce (van
der Aalst 1999, Lazcano et al. 2000) and virtual
enterprises (Casati et al. 1995, Liu and Pu 1998). In
manufacturing systems, Huang et al. (2000) proposes a
distributed workflow management model to develop
distributed manufacturing execution system. Lau et al.
(2000) present a methodology for the design and
development of a flexible workflow supply chain
(FWSC) system for achieving flexibility to cope with
unexpected changes. Chiu et al. (2001) use an integrated, event-driven approach for execution, coordination, and exception handling in workflow management
system. Montaldo (2004) applies workflow management system to enhance business performance for
small-medium enterprise. Tan and Fan (2007) adopt a
novel dynamic workflow model fragmentation algorithm to execute the distributed processes.
2.2. Agent-based manufacturing
Agent technology is a branch of artificial intelligence
(AI) and has been widely accepted and developed in


International Journal of Computer Integrated Manufacturing
manufacturing applications for its autonomy, flexibility, reconfigurability, and scalability (Sikora and Shaw
1998, Macchiaroli and Riemma 2002, Maturana et al.
2004). An agent based concurrent design environment
(Tan et al. 1996, Krothapalli and Deshmukh 1999) has
been proposed to integrate design, manufacturing and
shop-floor control activities. A compromising and
dynamic model in an agent-based environment (Sikora
and Shaw 1998) has been designed for all agents

carrying out their own tasks, sharing information, and
solving problems when conflicts occur. Some mobile
agent-based systems (Shin and Jung 2004) have been
applied to the real-time monitoring and information
exchange for manufacturing control. Jia et al. (2004)
proposed an architecture where many facilitator agents
coordinate the activities of manufacturing resources in
a parallel manner. Jiao et al. (2006) applied the MAS
paradigm for collaborative negotiation in a global
manufacturing supply chain network. Besides, in
various kinds of applications such as distributed
resource allocation (Bastos et al. 2005), online task
coordination and monitoring (Lee and Lau 1999,
Maturana et al. 2004), or supply chain negotiation
(Wu 2001), the agent-based approach has played an
important role to achieve outstanding performance
with agility.
2.3. Automatic data capturing for real-time wireless
manufacturing
Currently, real-time visibility and interoperability have
been considered core characteristics of next-generation
manufacturing systems (Huang et al. 2006). As early as
in early 1990s, Udoka (1992) has discussed the roles of
Auto ID as a real-time data capture tool in a computer
integrated manufacturing (CIM) environment. Early
RFID manufacturing applications have been briefly
quoted in Brewer et al. (1999) and further promoted in
Li et al. (2004). Johnson (2002) presents a RFID
application in a car production line. The website http://
www.productivitybyrfid.com/ also provides a few links

to real-life pilot cases. Chappell et al. (2003) provides
general overview on how Auto ID technology can be
applied in manufacturing. Pilot projects have recently
been implemented and reported ( archive/). Several relevant whitepapers have been prepared to provide roadmap for
developing and adopting Auto ID-based manufacturing
technologies (Harrison and McFarlane 2003, Chang
et al. 2003). More recently, the Cambridge Auto ID Lab
has launched an RFID in Manufacturing Special
Interest Group (SIG) ( Huang
et al. (2007, 2008) implement RFID technologies to
capture the real-time manufacturing data of employees,
machines and materials of assembly line.

3.

103

Overview of real-time reconfigurable manufacturing

The aim of the research reported here is to apply RFID
technologies and develop an easy-to-deploy and simpleto-use reconfigurable information infrastructure for
manufacturing companies to achieve real-time and
seamless dual-way connectivity and interoperability
between application systems at enterprise, shop-floor,
work-cells and RFID devices. Figure 1 shows an
overview of the proposed infrastructure. This infrastructure is consistent with a normal manufacturing hierarchy. That is, a manufacturing factory has one or more
shop-floor production lines. A production line consists of
several work-cells each of which involves a variety of
manufacturing objects, such as operators, machines,
materials etc. Different production lines are often

responsible for different production processes. According to the manufacturing hierarchy, the proposed RTM
infrastructure includes following core components:
. Shop-floor Gateway (SF-Gateway): SF-Gateway
is at the centre of the overall RTM. Following
Service-Oriented Architecture (SOA), SF-Gateway includes three main components, i.e. workflow management, MS-UDDI and work-cell
agent. These components will be further detailed
in the following sections.
. Work-cell Gateway (WC-Gateway): WC-Gateway
acts as a server to host and connect all RFIDenabled smart objects of a work-cell. A WCGateway has a hardware hub and a suite of
software systems. The hub physically connects all
RFID-enabled smart objects which are represented
as software agents in the WC-Gateway operating
system. All smart objects are used in a ‘universal
plug and play (UPnP)’ and interoperable manner.
. RFID-enabled smart objects: Smart objects are
those physical manufacturing objects that are
made ‘smart’ through equipping them with
RFID devices. Those with RFID readers are
active smart objects. Those with RFID tags are
passive smart objects. Smart objects interact with
each other through wired and/or wireless connections, creating what is called an intelligent
ambience. In addition, smart objects are also
equipped with specific operational logics, data
memory and processing functions. Therefore,
smart objects are able to sense, reason, act/react/
interact in the intelligent ambience community.
4. Overview of reconfigurable shop-floor gateway
The shop-floor gateway is a high-level key component
of the proposed real-time reconfigurable manufacturing infrastructure. Figure 2 shows the overall

104

Y. Zhang et al.

Figure 1.

RFID-enabled real-time manufacturing infrastructure.

Figure 2.

Overview architecture of reconfigurable shop-floor gateway.

architecture of the reconfigurable SF-Gateway. Following the service-oriented architecture (SOA), workcells, work-cell applications, equipments and smart
object in a shop-floor can all be considered as
manufacturing services. The manufacturing services
(agents) are wrapped as a work-cell agent. Each agent
contains the real-time information and status of the
work-cell. Work-cell agents can be registered and
published at MS-UDDI while SF-Gateway can select,
add and deploy these work-cell agents for different

production process, enabling a reconfigurable manufacturing (Huang et al., 2008a). Specifically, process
planners use configuration facilities to search suitable
work-cell agents from MS-UDDI and configure them
for a specific production process. The configuration
result is a workflow among work-cell agents (i.e. workcells) which represents a manufacturing process. This
workflow could be used for the next lifecycle stage –
execution. During the actual execution, shop-floor
managers could monitor and control the status and

International Journal of Computer Integrated Manufacturing
progress of the manufacturing process through SFGateway tools which communicate with work-cell
agents at gateway servers. Real-time data are handled
centrally at the SF-Gateway repository.
The proposed reconfigurable SF-Gateway is composed of three major components, namely workflow
management, UDDI for manufacturing services (MSUDDI), and work-cell agents.
4.1.

Workflow management (WM)

This WM component is used to (1) define a manufacturing workflow based on a product’s process plan,
(2) (re)configure the manufacturing services through
using agent-based workflow model and MS-UDDI,
and (3) coordinate the involved work-cell agents to
execute real-time manufacturing according to the
defined workflow. Workflow management includes
three modules, namely definition module, binding
module and execution engine. They are described as
following.
. Definition module is responsible for defining the
workflow according to the specific production
processes.
. Bind module automatically records the relationships between the work cell agents and the
manufacturing process after work-cell agents are
searched and chosen as production process nodes.
. Execution engine not only facilities the execution
of work-cell agents according to the defined
workflow and logic, but also monitor and control

the work-cell agents during the execution
process.
4.2. Manufacturing services UDDI (MS-UDDI)
The function of MS-UDDI is similar to those of
standard UDDI (Universal Description, Discovery
and Integration), serving as a platform-independent
framework for describing and discovering services
through Internet. MS-UDDI is composed of three
main modules, i.e. publish and search module, service
model and tModel.
. Publish and search module is used to convert
work-cell agents into public web services, which
can be easily searched to implement flexible and
reconfigurable shop-floor manufacturing.
. Service model describes a group of web services
which are contained in a businessService structure. For each published work-cell agent, there is
a set of web services each of which serves for
certain specific propose and can be invoked over
internet.

105

. tModel is a data structure representing a service
type (a generic representation of a registered
service) in UDDI.
4.3. Work-cell agent
Work-cell agent is responsible for wrapping the workcell applications to process the complex real-time data
captured from smart objects. The work-cell agent can
be used to reflect the real-time manufacturing information and status of a work cell. The work-cell agents
must be registered and published at MS-UDDI as web

services, then found and invoked via MS-UDDI so
that the proposed reconfigurable manufacturing can be
achieved. Along this innovative concept and structure,
other users or systems can directly attain the status and
real-time information of work cell by visiting or
invoking its agents. Three models, method model,
data model and smart object manager (SOM) model,
are involved in this component.
. Method model is used to deal with the huge realtime data captured by RFID devices installed at
the work-cell Gateway based on rules and
schemas. For example, the ‘getMaterials’ method
will deal with all the data relevant to materials of
this work-cell Gateway and return detailed realtime information such as material item, quantity
etc. At MS-UDDI, each method can be regarded
as a service of the registered work-cell agent.
. Data model describes and defines the basic
standards of input and output data of work-cell
agents. The data model adopts XML-based
schema that can be easily edited, transformed
and extended.
. Smart object manager model aims at managing
the behaviors of smart objects installed at a
work-cell gateway. It is implemented with
intelligence logics so as to sense and identify
the real-time manufacturing data of single smart
agents as well as the communication between
multiple smart objects.
5. Agent-based workflow management for
reconfigurable manufacturing
5.1. Agent-based workflow management model for

reconfigurable manufacturing
An agent-based workflow management (WFM) model
will be firstly defined of its topology of processes and
manufacturing resources, i.e. indicates the involved
manufacturing resource types yet has not been assigned
to specific work cells (agents). Then, each node of the
workflow will choose its execution agent from MSUDDI based on the actual status of the qualified agents.


106

Y. Zhang et al.

Figure 3 shows the agent-based WFM framework,
which is used to not only plan and control the flow of
production processes and data, but also executes any
process (work) node of the workflow from the selected
work-cell agent. The WFM model is responsible for
reusing and reconfiguring the work-cell agents to
implement various production orders. There are two
basic elements in the agent-based WFM model:
process and agent. A process refers to a portion of a
production task and can be assigned to a specific
manufacturing resource (work cell). As mentioned in
section 3, the work-cell agent wraps the corresponding
function of a work-cell, which can execute and finish
the process.
In the SF-Gateway level, the WFM is mainly
concerned with the co-ordination of distributed workcell agents. As can be seen in Figure 3, process nodes
represent production tasks, and the logical nodes

represent the trigger conditions. Edges represent the
logical relationships between production processes, i.e.
the flows of control and data.
WFM is built on the concept of agents. An agent
represents a work package in the workflow. All the
agents involved in a workflow share the same
repository and the repository becomes a common
working memory. This sharing information ensures the
traceability of the decisions at different stages by
recording them in a decision tree in terms of the
contents of the decisions, the decision-makers, and
precedence decisions, etc.
Agents are only used to define the work or node of
a production process workflow. Relationships between
nodes are separately defined in terms of flows. Without
flow definitions, agents still do not know where inputs

Figure 3.

Agent-based workflow management model.

are obtained from and outputs are sent to. The
separation of flow definition from work definition
provides opportunities to reuse agents for different
production projects once they are defined for RTM.
No further changes are necessary when agents are used
for other production projects. What the project team
needs to do is to choose the agents according to the
different production project and define the flows of
control and data between agents to suit specific

requirements.
5.2. Workflow definition
Workflow definition has two work modes. One is
editing mode which means the process planner defines
the agent-based workflow for a specific production
project. The other is executing mode which means the
manager monitors and controls the progress of
executing a production workflow. Workflow definition
in turn involves the ‘work’ definition and the ‘flow’
definition.
Two types of flow are identified in this WFM. They
are flow of precedence and flow of data. The flow of
precedence and logic node between work-cell agents
defines their dependencies. For example, supposing a
simple hypothetical product consists of two components, B and C. The component B is an outsourcing
and the component C is produced at work-cell 2.
Finally, component B and C are assembled to form
Product A at work-cell 3. Accordingly, the production
of A is decomposed into three production processes
which can be depicted by the directional networktopology mode. Here, Agent 1 represents a ‘delivery A’
work, Agent 2 represents a ‘producing B’ work and


International Journal of Computer Integrated Manufacturing
Agent 3 represents an ‘assembling C’ work. As shown
in Figure 4, Agent 3 can only start its work after Agent
1 and 2 complete their works under the and logical
condition. Agents 1 and 2 may work simultaneously.
The flow of data refers to the situation where
agents share their property data. Some outputs from

an agent may be the inputs to other agents. Such
relationships can be easily defined in a similar way that
relationships are defined between data tables in a
relational database. Flows of data can be compared to
messages widely used in multi-agents system (MAS)
for communication. And the message configuration
tool configures where inputs are obtained from and
outputs are send. E Figure 4 combines some output
items of Agents 1 and 2 as the one input item of Agent
3 according to the real requirements.

Figure 4.

Two types flow of work: control flow and data flow.

Figure 5.

Agent-based execution of a workflow model.

107

Flows of data, or message passing, are triggered by
the flow of precedence and logical condition. For
example, during the ‘and’ condition, if Agents 1 and 2
have not finished with its work, flows of data
associated with Agent 3 will not be processed.
5.3.

Workflow execution

Once a workflow is fully defined, it can be executed as
illustrated in Figure 5. During the execution, each node
in the workflow will be translated to an agent. Each
agent will invoke its manufacturing services (e.g. workcells, smart objects, etc.) of the real manufacturing
environment to enable their intelligent management of
the manufacturing process. Explorers are provided to
operators, managers and supervisors for monitoring


108

Y. Zhang et al.

and controlling the workflow execution lifecycle. The
users can simply follow the logic and execute relevant
production tasks. As a high-level user, the shop-floor
manager can have a clear overview of the progress of a
production project at the SF-Gateway. At the WCGateway, on the other hand, the operators of workcells can use this facility to check if the conditions of
their tasks are met so that they can start. The general
procedure of executing a workflow is as follows:
. The work-cell Agents use Service Oriented
Architecture framework to connect to the web
server where SF-Gateway is deployed;
. XML-Based workflow definition file is automatically downloaded to and manually activated at
the corresponding Agents;
. Repository is contacted to retrieve the workflow
model defined in advance;
. The first agent in the workflow is activated. The
agent is executed according to the procedure
discussed in the preceding section. Its incoming

messages defined as flows of data associated are
fired. Therefore, this agent knows from where its
input data come;
. After preparing its input data, repository is
contacted to save the input/output and other
data of the agents;
. Execution engine notifies all work-cell agents
about the changes;
. The work-cell agent is prompted if the output is
accepted or a backtracking is necessary;
. Upon completion, the control is passed over to
subsequent agents; and

Figure 6.

Overview of the motivating assembly line.

. This process repeats until the last agent in the
workflow is completed.

6. Case study
6.1. Configuration of a representative assembly line
This section demonstrates the usage of the proposed
architecture with an example application of an
assembly line. This proof-of-the-concept test-bed
serves the purpose of demonstrating how the proposed
real-time reconfigurable WM framework would work
in an industrial environment, gaining insights about
requirements of WM solutions, and highlighting
further issues for research and development of WM

solutions. The study is based on a simplified motivating scenario shown in Figure 6.
The configuration of the assembly line and workcells depends upon the structure of the product
(variant) to be assembled. The product demonstrated
here is composed of five components assembled
sequentially across three work-cells, as shown in
Figure 6. At the first work-cell, two components,
one of which is a critical base module, are put
together to form the first level subassembly. Two
further components are added to the subassembly at
the second work-cell. The last component is assembled to produce the finished product at the third
work-cell where an inspection also takes place to
determine if the product is acceptable or rejected. All
components and subassemblies are moved and maintained in containers or pallets of appropriate sizes
and shapes.


International Journal of Computer Integrated Manufacturing
6.2.

Agent publish

In order to implement the proposed agent-based
workflow management for shop-floor, work-cell agents
should be firstly registered and published to MSUDDI. Figure 7 illustrates the main steps for publishing a work-cell agent at MS-UDDI, namely
Step 1: Login MS-UDDI
The manufacturing resource manager is responsible
for registering and publishing the work-cell agents of
as web services. Considering the security, the manufacturing resource manager needs to login the MSUDDI using his account for further operations as seen
in Figure 7(a).
Step 2: Register Service information of Work-cell

Agent
The MS-UDDI provides facilities, as shown in
Figure 7(b), to describe the service information of the
work-cell agent so that users (e.g. shop-floor manager) can easily know what services this work-cell can
provide. The registered information includes agent
ID, description, process capability, access point etc.
Agent ID is used to identify the each work-cell
agent; description indicates the machine type of the
work-cell agent, e.g. lather, mill, assembly etc.;
process capability describes the specific machining
characters, i.e. which processes can be executed at
this work-cell; access point shows the internet address
of a work-cell agent to enable users or systems to get
access.
Step 3: Publish Work-cell Agent
After the registration of agent information, the
manufacturing resource manager can physically
publish the work-cell agents to MS-UDDI, as can
be seen in Figure 7(c). A published work-cell agent
can be found and invoked as web service through
internet. This register and publish processes are
repeated until all the work-cell agents are published
to MS-UDDI.

Figure 7.

6.3.

109

Workflow definition

After all the work-cell agents have been defined and
published at MS-UDDI, it is important to establish the
workflow of the shop-floor level. The definition of a
shop-floor workflow includes four steps as can be seen
in Figure 8. The main steps are described as following:
Step 1: Get the process planning of the specific
product
A shop-floor workflow is corresponding to a
practical production process. Generally, different
products have different production processes. Therefore, process planning of a product is the input
information for defining a shop-floor workflow. The
detailed process planning of a product ‘A’ is given at
the top of Figure 8, including three assembly processes,
i.e. Assembly ‘C’, ‘B’ and ‘A’.
Step 2: Define workflow according to the process
planning
The workflow facility provides graphic interfaces
for shop-floor manager to edit the work and flow
according to specific process planning. As seen in
Figure 8(b), the three production processes of product
‘A’ are described by graphic objects. The rectangle
denotes the ‘Assembly’ work, and the directed arrow
represents the sequence of the processes.
Step 3: Establish mappings between the processes
and agents
This step is responsible for establishing mapping
relationships between processes and agents, as shown
in Figure 8(c). For each process, e.g. the process

‘Assembly B’, there are several potential work-cell
agents registered at the MS-UDDI. The optimal agent
can be selected to execute a process based on its
capabilities, e.g. capability and its real-time status.
Then, a pop-up dialog is used to define the condition
constraints and input/output data. At this time, the
mapping relationship between the process and agent is
established.

Main steps for publishing work-cell agent at MS-UDDI.


110

Y. Zhang et al.

When the workflow operation is carried out, an
XML segment is created simultaneously. The XML
segment stores the condition parameters and input/
output data, etc. defined by user.
Step 4: Create workflow definition file
The step 3 is repeated until all the condition and
data of all the processes and its agents have been
defined, and the corresponding XML-based workflow
definition file is formed at this time, as shown in Figure
8(d). So far, the definition of agent-based workflow has
been finished.

6.4.

Workflow execution and monitor

The defined workflow definition file will be imported to
the workflow engine for execution through coordinating the involved work-cell agents. During the execution
process, agent explorer and workflow explorer could
be used for tracing and monitoring the execution status
of work-cell agents and workflow respectively. Figure 9
shows the two explorers.
Once an agent is started by the workflow engine,
the agent explorer is available for viewing the real-time

Figure 8.

Main steps for definition an agent-based workflow.

Figure 9.

Execution and monitoring explorer. Available in colour online.


International Journal of Computer Integrated Manufacturing
information of this work-cell, as shown in Figure 9(a).
Production operators can use this useful and real-time
manufacturing information to execute assembly plans
and schedules at the level of individual work-cell.
Figure 9(b) shows an overall execution of all the
involved work-cell agents. Each work node represents
an agent whose real-time status could be tracked. In
this case, different statuses of work-cell agents are
indicated by three different colours. Red status

indicates that the assigned work of a work-cell agent
has not been started; yellow status indicates the
assigned work is being processed; blue status indicates
the assigned work has been finished. This explorer
provides real-time production information for high
level administrators of enterprise to improve shopfloor productivity and enhance the rapid responsibility
in case of onsite exception.
7. Concluding discussion
The shop-floor gateway presented in this paper is an
easy-to-deploy and simple-to-use infrastructure framework which integrates the concept of agents into
workflow management and RFID devices to realise
real-time reconfigurable wireless manufacturing. On
the one hand, RFID technologies are used to achieve
real-time manufacturing data collection, and enable
the dual-way connectivity and interoperability between
high-level (i.e. shop-floor level) and work-cell level,
and create real-time visibility and traceability throughout the entire enterprise. On the other hand, agentbased workflow management model is applied to plan,
reconfigure and monitor the flow of production
processes, data and control, and manage the realtime manufacturing information. The use of this
proposed technology will allow manufacturing enterprises to improve shop-floor productivity and quality,
reduce the wastes of manufacturing resources, cut the
costs in manufacturing logistics, reduce the risk and
improve the efficiency in cross-border customs logistics
and online supervision, and improve the responsiveness to market and engineering changes.
Three contributions are important in our research.
The first contribution is the integration of the agent
concept into workflow management. Production processes at the shop-floor level are represented as
workflows. At the workflow definition stage, the
shop-floor manager achieves real-time reconfigurable
manufacturing by configuring work-cell agents for a

production process. At the workflow execution stage,
real-time manufacturing is achieved by capturing and
collecting the real-time data of smart objects installed
at all work-cells. The second contribution is the use of
agents based on service-oriented architecture to wrap
the work-cell applications as real-time manufacturing

111

information services. Therefore, RFID-enabled manufacturing objects can be easily reused and reconfigured
to form specific production processes from agent-based
services registered at MS-UDDI. The third contribution is the overall framework for the RFID-enabled
real-time manufacturing infrastructure. Through installing the RFID devices including RFID readers and
tags to traditional manufacturing resources such as
employees, equipments and materials, the real-time
manufacturing data of shop-floor can be cost-effectively captured and collected to support dynamic
decision-making activities.
However, the work should be further extended in
several ways. For example, the design and development of the micro workflow management of the smart
object to manage the whole lifecycle of manufacturing
services at the level of wireless devices. Micro workflow
management facilities facilitate smart objects so that
they can sense, identify, communicate, decide and act.
In addition, other enterprise application systems (e.g.
Advanced Planning and Scheduling - APS and
Computer Aided Process Planning - CAPP) should
be considered as inputs to the proposed workflow
management. Most suitable work-cell agents (services)
can then be selected based on the results of APS and
CAPP services. The real-time manufacturing information captured by work-cell agents can be used to

enhance the real-time decisions at shop-floor and
enterprise levels.
Acknowledgements
The authors are most grateful to various companies who
provide technical and financial supports to this research. The
authors would like to acknowledge financial supports of
HKSAR ITF (GHP/042/07LP) grant, HKSAR GRF (HKU/
712508), National Science Foundation of China (50805116;
70629002), HKU Research Committee Grants, and from
industrial collaborators.

References
Bastos, R.M., Oliveira, F.M., and Oliveira, J.P., 2005.
Autonomic computing approach for resource allocation.
Expert Systems with Applications, 28 (1), 9–19.
Brewer, A., Sloan, N., and Landers, T., 1999. Intelligent
tracking in manufacturing. Journal of Intelligent Manufacturing, 10 (3–4), 245–250.
Casati, F., et al., 1995. Conceptual modelling of workflows.
In: Proceedings of the International Conference on ObjectOriented and Entity-Relationship. Berlin: Springer, 341–
354.
Chang, Y., McFarlane, D., and Koh, R., 2003. Whitepaper
on ‘Methodologies for integrating Auto ID data with
existing manufacturing business information systems.
Available from: />publications/w_papers/cam-autoid-wh009.pdf
Chappell, G., et al., 2003. Auto ID on the line: the value of
auto ID technology in manufacturing. Auto ID Center,
CAN-AutoID-BC-005.


112

Y. Zhang et al.

Chiu, D.K.W., Li, Q., and Karlapalem, K., 2001. Web
interface-driven cooperative exception handling in
ADOME workflow management system. Information
Systems, 26 (2), 93–120.
Elmagarmid, A. and Du, W., 1998. Workflow management:
state of the art versus state of the products. In: A. Dogac,
L. Kalinichenko, M. Tamer Ozsu, and A. Shethm, eds.
Workflow Management Systems and Interoperability.
NATO ASI Series, Series F: Computer and Systems
Sciences 164, Springer Heidelberg, 1998.
Harrison, M. and McFarlane, D., 2003. Whitepaper on
development of a prototype PML server for an auto
ID enabled robotic manufacturing environment. Available from: />publications/w_papers/cam-autoid-wh010.pdf
Hollingsworth, D., 1994. Workflow management coalition
the workflow reference model. Workflow Management
Coalition. Document Number TC00-1003, UK.
Huang, C.Y., 2002. Distributed manufacturing execution
systems: a workflow perspective. Journal of Intelligent
Manufacturing, 13 (6), 485–497.
Huang, G.Q., et al., 2008a. RFID-enabled smart objects for
real-time reconfigurable manufacturing. Proceedings of
DET2008 5th International Conference on Digital Enterprise Technology, 22–24 October 2008, Nantes,
France.
Huang, G.Q., et al., 2008b. RFID-based wireless manufacturing for real-time management of job shop WIP
inventories. International Journal of Advanced Manufacturing Technology, 23 (4), 469–477.
Huang, G.Q., Huang, J., and Mak, K.L. 2000. Agent-based
workflow management in collaborative product development on the Internet. Computer-Aided Design, 32, 133–

144.
Huang, G.Q., Wright, P.K., and Newman, S.T., 2009.
Wireless manufacturing: a literature review, recent
development, and case studies. International Journal of
Computer Integrated Manufacturing, 22 (7), 1–16.
Huang, G.Q., Zhang, Y.F., and Jiang, P.Y., 2006. RFIDbased wireless manufacturing for walking-worker assembly shops with fixed-position layouts. International
Journal of Robotics and Computer Integrated Manufacturing, 23 (4), 469–477.
Jia, H.Z., et al., 2004. An adaptive upgradable agent-based
system for collaborative product design and manufacture. Robotics and Computer-Integrated Manufacturing,
20 (2), 79–90.
Jiao, J.R., You, X., and Kumar, A., 2006. An agent-based
framework for collaborative negotiation in the global
manufacturing supply chain network. Robotics and
Computer Integrated Manufacturing, 22 (3), 239–255.
Johnson, D., 2002. RFID tags improve tracking, quality on
Ford line in Mexico. Control Engineering, 49 (11), 16–16.
Koren, Y., et al., 1999. Reconfigurable manufacturing
systems. Annals of the CIRP, 48 (2), 527–540.
Krothapalli, N. and Deshmukh, A., 1999. Design of
negotiation protocols for multi-agent manufacturing
systems. International Journal of Production Research,
37 (7), 1601–1624.
Lau, H.C.W., Ning, A., and Chan, F.T.S., 2000. Design and
development of a flexible workflow supply chain system.
IEEE International Conference on Management of Innovation and Technology Proceedings, 12–15 November
2000, Singapore, 792–796.
Lazcano, A., et al., 2000. WISE: process based e-commerce.
IEEE Data Engineering Bulletin, 46–51.

Lee, W.B. and Lau, H.C.W., 1999. Multi-agent modelling of

dispersed manufacturing networks. Expert Systems with
Applications, 16 (3), 297–306.
Li, Z.K., Gadh, R., and Prabhu, B.S., 2004. Applications of
RFID technology and smart parts in manufacturing.
Proceedings of DETC’04: ASME 2004 Design Engineering Technical Conferences and Computers and Information
in Engineering Conference, September 28–October 2,
2004, Salt Lake City. Utah USA DETC2004–57662.
Liu, L. and Pu, C., 1998. Methodical restructuring of
complex workflow activities. Proceedings of the ICDE
1998. IEEE Computer Society Press, 342–350.
Macchiaroli, R. and Riemma, S., 2002. A negotiation scheme
for autonomous agents in job shop scheduling. International Journal of Computer Integrated Manufacturing, 15
(3), 222–232.
Maturana, F.P., et al., 2004. Distributed multi–agent
architecture for automation systems. Expert Systems
with Applications, 26 (1), 49–56.
Mehrabi, M.G., Ulsoy, A.G., and Koren, Y., 2000. Reconfigurable manufacturing systems: key to future manufacturing. Journal of Intelligent Manufacturing, 11, 413–419.
Montaldo, E., Sacile, R., and Boccalatte, A., 2003. Enhancing workflow management in the manufacturing information system of a small–medium enterprise: an
agent-based approach. Information Systems Frontiers, 5
(2), 195–205.
Radunovic, B., 1999. An overview of advances in reconfigurable computing systems. Proceedings of the 32nd
Hawaii International Conference on System Sciences,
IEEE, 1–10.
Schal, T., 1996. Workflow management systems for process
organizations. Lecture Notes in Computer Science. New
York: Springer-Verlag.
Shin, M. and Jung, M., 2004. MANPro: mobile agent-based
negotiation process for distributed intelligent manufacturing. International Journal of Production Research, 42
(2), 303–320.
Sikora, R. and Shaw, M.J., 1998. A multi-agent framework

for the coordination and integration of information
systems. Management Science, 44 (11), 65–78.
Tan, G.W., Hayes, C.C., and Shaw, M., 1996. An intelligentagent framework for concurrent product design and
planning. IEEE Transactions on Engineering Manufacturing Management, 43 (3), 297–306.
Tan, W. and Fan, Y.S., 2007. Dynamic workflow model
fragmentation for distributed execution. Computers in
Industry, 58 (5), 381–391.
Udoka, S.J., 1992. The role of automatic identification (Auto
ID. in the computer integrated manufacturing (CIM
Architecture). Computers and Industrial Engineering, 23
(1–4), 1–5.
Van der Aalst, W.M.P., 1999. Process-oriented architectures
for electronic commerce and inter organizational workflow. Information Systems, 24 (8), 639–671.
Wu, D.J., 2001. Software agents for knowledge management:
coordination in multi-agent supply chains and auctions.
Expert Systems with Applications, 20 (1), 51–64.
Yigit, A.S. and Usloy, A.G., 2002. Dynamic stiffness
evaluation for reconfigurable machine tools including
weakly non-linear joint characteristics. Proceedings of the
Institute of Mechanical Engineers, Part B, 216, 87–100.
Zhao, X.B., Wang, J.C., and Luo, Z.B., 2000. A stochastic
model of a reconfigurable manufacturing system, Part 1:
a framework. International Journal of Production Research, 38 (10), 2273–2285.


International Journal of Computer Integrated Manufacturing
Vol. 23, No. 2, February 2010, 113–125

Internet-based intelligent service-oriented system architecture for collaborative product
development

Yi Hu*, Xionghui Zhou and Congxing Li
National Die & Mold CAD Eng. Research Center, Shanghai Jiao Tong University, Shanghai, 200030, China
(Received 31 October 2008; final version received 27 October 2009)
Collaborative product development is a strategic approach for manufacturing enterprises to succeed in the
competitive market. Issues of information and knowledge sharing within the value chain are regarded as the most
significant aspect of collaborative product development and the sharing is based on geographically distributed
business domains. To support this issue, this paper discusses internet-based intelligent system architecture for
collaborative product development built upon service-oriented architecture that is a profitable infrastructure for
handling distributed heterogeneous resource sharing. The system utilises basic and extensible services to provide
efficient and effective knowledge-sharing functionality on-demand. A Java-enabled solution together with artificial
intelligence techniques is employed to develop such a networked system. Finally, a case study is presented to
illustrate how the tool and die makers are involved and cooperate with the product designers to make an optimal
design collaboratively.
Keywords: knowledge-sharing; service-oriented architecture; collaborative product development; tool and die maker

1.

Introduction

Today, the world has become more and more practical
and cost-conscious. This trend has dramatically
intensified and globally expanded the competitive
environment of manufacturing in recent years. These
developments require that the process of product
development continuously evolves so as to satisfy
consumer demands in a changing market and shrink
product life cycles. Enterprises must boost their flexibility and responsiveness in terms of product development so as to design and produce more complex
products at lower cost and in less time. In this global
environment, organisations do not possess all the
knowledge they need but instead rely on buying technologies or services through contractual and cooperative partnerships with other organisations (Choo et al.

2000). As a result, more and more efforts focus on
building up a new product development paradigm that
requires various domain experts to collaborate closely
and share knowledge among themselves. The degree
of knowledge sharing supported by a collaborative
environment is an important parameter for its success
in the global market. In response to this need, a solution called collaborative product development (CPD),
which is defined as: ‘an internet based computational
architecture that supports the sharing and transferring
of knowledge and information of the product lifecycle
amongst geographically distributed companies to aid

*Corresponding author. Email:
ISSN 0951-192X print/ISSN 1362-3052 online
Ó 2010 Taylor & Francis
DOI: 10.1080/09511920903440347


taking right engineering decisions in a collaborative
environment’’ (Rodriguez and Al-Ashaab 2002), has
been proposed by research community. However, the
lack of communication and interaction among partners often results in long development time, high cost,
incompatible problems and other inconsistencies.
Therefore, information technologies have been applied
to develop a CPD system recently to enhance their
ability to share knowledge and work collaboratively
during the whole product development lifecycle. As
different roles in the product development lifecycle
are always with different knowledge and different
levels of expertise, the intelligent collaboration among

different roles appears to be important.
In most cases, intelligent collaboration always
covers several different domains and platforms that
may be heterogeneous or homogenous, structured
or unstructured, distributed or centralised so that
traditional technologies are limited. To cope with this
problem, the approach of distributed computing platform based knowledge-sharing, integrated with
eXtensible Markup Language (XML), becomes the
most likely direction. XML is a standard for data representation and exchange on the Internet and provides
a common way for defining a specific format of message
that can be transported in heterogeneous environment
without misunderstanding. A special advantage of
XML is standard application programming interfaces


114

Y. Hu et al.

(APIs) for parsing XML documents, so XML users do
not have to develop their own parsers for the specific
XML formats they use (Jovanovic and Gasevic 2005).
There are several approaches that propose XML as the
common syntax for sharing knowledge. For example,
Leff (Leff 2001) developed Jess User Functions that
load XML documents and convert their document
object model (DOM) tree into series of facts that can be
consumed by Java expert system shell (Jess). On
another side, service-oriented architecture, as one of
distributed computing platforms, exposes concrete

functionality as standards-based, shared and reusable
services. This architecture is a collection of declared
services that are independent and loosely coupled, but
controlled through policies. Service-oriented architecture provides an infrastructure to support knowledgesharing in geographically distributed heterogeneous
environments.
This paper describes the internet-based intelligent
service-oriented system architecture for collaborative
product development in order to propose an enabler
tool used in collaborative product development and
discusses the holistic architecture, methodology and
prototype implementation. The remainder of the paper
is organised as follows. Section 2 presents the background of collaborative product development with
related research review and service-oriented architecture. Section 3 puts forward the detailed description of
the system architecture and the methodology adopted.
Section 4 analyses the prototype implementation.
Section 5 concludes the paper with future research
work.
2.

Background and related research review

2.1. Collaborative product development
Products born in the twentieth century have increased
in complexity, which implies complex product development processes, complex organisational forms and
more complex projects for managing the process of
developing the products (Williams 1999). Twentieth
century globalisation has also brought increased competition from various parts of the world. Both the
increased complexity of products and the increased
competition on global market imply a need for
collaboration in product development.

Collaborative product development (CPD) concepts are not new and revolutionary. The theoretical
concept of CPD first started to appear in 1994, in a
journal paper ‘Multimedia Comes of Age’ (Cassidy
1994). In 1995, the CPD concept appeared in many
more publications by authors such as Bruce et al.
(1995). At that time, the research of CPD was around
the buyer–supplier relationships, as well as the complexity and success factors of CPD. Later on, CPD

appeared frequently in journal articles and conference
papers. CPD focuses on the continued responsibility of
different disciplines and lifecycle functions for product
development specifications and their translation into a
product that satisfies the customer. Thus, during the
lifecycle of CPD, every stage is an iterative process and
includes some sub-processes, tasks, and activities. For
example, during product definition, collaborative design, collaborative simulation and product evaluation
are involved in the product design stage, which takes
time and is costly. This in turn demands a more holistic
view and supporting tools, in the form of methods,
models, or processes.
A number of research initiatives related to CPD
have been undertaken by several authors before.
Several technological requirements that must be
addressed in order to develop enabling technologies
for CPD are summarised below:
(1) System architecture: To have a holistic view of
the CPD system architecture, models and applications should be integrated within a framework
in a structured and transparent manner. Appropriate system architectures should be designed to
support collaborations and coordinate CPD
activities effectively and efficiently.

(2) Communication and collaboration: During the
product development, there are always many
kinds of changes, such as a design requirement,
an unanticipated simulation or testing result, the
availability of a component, or an improvement
to the manufacturing process. It is essential for
quality and productivity to react quickly to such
changes. Knowledge management, knowledgesharing mechanisms, the documentation of
learning lessons and other generic rules are
also the concerns of this requirement.
(3) Model and management: The models of product and process are the basic features of
CPD. The product model should support and
facilitate the integration of every phase of
CPD process, like between design and simulation, testing and evaluation, i.e., the integration
between different CAD models, between CAD
models and CAE/CAM models, etc. Managing
a CPD model and process is much more
complex.
(4) Implementation methodologies: Suitable CPD
methodologies are helpful and critical for better
understanding and successful implementation
of CPD and provide enabling tools for CPD.
To develop a CPD system meeting all the above
requirements, it is necessary to have a distributed,
platform-independent and intelligent architecture


International Journal of Computer Integrated Manufacturing
based on new information technologies to cope with
these challenges.

2.2. Related research
Authors from various areas have addressed collaborative product development in different ways.
Integrated product models contain all product data
needed during product lifecycles in order to support
collaborative development. Many works have been
done before, such as FBS model (Gero 1990), Krause
et al. (1991), Tichkiewitch et al. (1993), the CPM
model (Sudarsan, et al. 2004, Noel et al. 2004), the
IPPOP intiative (Roucoules, et al. 2006) and the PPO
design model (Noel et al. 2008).
Research on information technologies applied in
CPD is rather extensive. Examples include Huang,
et al. (2000), Kumar and Midha (2001), Toerlind
(1999), Ramesh and Tiwana (1999) and OehrwallRoennbaeck (2002). Web based technologies (including Java, JSP, ASP, servlet and XML) have been
widely used in the implementation of CPD systems
(Roy et al. 1991, Qiang et al. 2001, Shen et al. 2001,
Tay and Ming 2001, Zha and Du 2002, Mok et al.
2008). Distributed object technologies (including
CORBA, COM/DCOM/COMþ, and Java RMI)
have been used in commercial software to develop
sophisticated solutions for the industry. Research
on agent-based systems within the area of artificial
intelligence (AI) has also emerged during the last
decades (Shen and Wang 2003, Dunbing Tang 2004).
Apparently, research on computer supported collaborative work is well covered and is now the focus of
many established research groups.
Decision support systems (DSS) are used in virtual
collaborations between organisations (e.g. virtual
enterprises (VE)) or within an organisation among
organisational functions (Huang et al. 2000) (e.g. CE

or IPD) to support the everyday work and facilitate the
decision-making.
Content management (Gsell 2006) is an additional
emerging area and unquestionably an important
part of a collaborative product development. Systems
that manage product content or data have been
referred to as product data management systems
(PDM). In related PDM application projects, commercial PDM tools (SmartTeam, Motiva) are used to
implement enterprise integration. Similarly, a generic
template for CPD is presented (Zhang et al. 1999),
which adopts PDM as its universal integration
platform.
Supply chain issues have been extensively addressed within the purchasing area (Fraser et al.
2003). However, recently, the focus of supplier
collaboration has also been addressed from the

115

product development perspective (e.g. Peter 1996,
Culley et al. 1999, Handfield et al. 1999, Wynstra
et al. 2001, Fagerstroem and Jackson 2002).
A distributed collaborative environment is required
to achieve concurrent and cooperative product development. Recently emerged web services, semantic web,
and grid computing technologies are promising in
the implementation of more flexible and scalable
CPD systems (Chung et al. 2000, Beiter and Ishii
2003, Johanson, and Krus 2003). As the SOA with
Web service is inherently distributed and scalable, a
SOA-based collaborative environment with knowledge-sharing for product development is developed in
this research to enable the inter-communications,

negotiations, coordination and cooperation between
the participants for joint and concurrent product
development.
2.3. Service oriented architecture
In recent years, service-oriented architecture (SOA)
has emerged as a key strategy for improving information technologies performance. In fact, Gartner, Inc.
(the world’s leading information technology research
and advisory company) projects that by 2010 SOA
will be the design standard in more than 80% of new
mission-critical applications and processes. SOA is
becoming a leading paradigm for information planning
and application integration.
Adopting the practice of SOA delivers important
benefits as follows:
(a) Faster time-to-market for new applications
helps organisations achieve competitive advantage by rapidly seizing new opportunities and
responding to threats.
(b) Flexible applications create flexible processes,
making the organisation more agile and adaptable in the face of changing market
requirements.
(c) Service reuse creates greater efficiency and
lowers maintenance costs.
(d) The ability to respond quickly to new regulatory requirements or specific compliance issues
helps companies avoid governmental penalties.
Thereby, SOA dramatically improves the flexibility
and adaptability of enterprises by accelerating the
time to market for new applications and processes,
driving down costs by making services highly reusable
and building processes to support changes. As a result,
enterprises can react faster, seize new opportunities

and respond more quickly to competitive threats,
which are the same requirements of collaborative
product development and have shown great


116

Y. Hu et al.

potential in distributed and collaborative design and
manufacturing.
Until now, little research has been found to apply
the SOA approach into collaborative product development, which still deserves extensive research. In this
paper, intelligent SOA system architecture is constructed to integrate die-maker’s activities into customer product development process in the distributed
and collaborative environment. It has functions to
assist participants to communicate easily and conveniently and to be aware of the mutual requirements
during the early design stage.
3. Internet-based intelligent service-oriented system for
CPD
3.1. Overall system architecture
The internet-based intelligent service-oriented system
architecture is depicted in Figure 1. In general, the
system architecture can be divided into six main
components: Base Resource, Integration, Service Repository/Registry, Orchestration, Governance & Management and Presentation.
The Base Resource is the physical environment that
consists of packaged or internally developed applications and data sources, such as a knowledge base,
database or other legacy applications. It is the foundation of the system and provides the information
and capabilities of various entities for the upper
components to repackage and expose as reusable
services, so that services can meet functionality,

performance and availability requirements.

Figure 1.

The Integration plays as an intermediary between
the underlying Base Resource and the upper components that need to access and share resources in the
system. Integration enables the efficient and flexible
utilisation of resources across and beyond the boundaries of organisations and enables them to interoperate
seamlessly, despite the existence of multiple and
possibly heterogeneous platforms and protocols, while
leveraging the potential of the Internet. It typically
includes integration technologies, such as data integration, enterprise services bus (ESB), Enterprise application integration (EIA), enterprise information
integration (EII), etc.
The Service Repository/Registry consists of reusable and self-contained services and information
or meta-date of them, providing the capability of
registering, searching and finding services. A service is
provided by an entity for use by others. In this system,
a service is created to enable access to one or more
capabilities in the Base Resource through the Integration. A service is accessed by means of a service
interface, where the interface comprises the specifics
of how to access the underlying capabilities. There are
no constraints on what constitutes the underlying
capability or how access is implemented. Thus, the
service could carry out its described functionality
through one or more processes and they could invoke
other available services. It represents the new paradigm
for commercial software development and delivery and
is often prominent in SOA strategy.
The Orchestration provides tools for creating
applications, defining workflows and assembling

Internet-based intelligent service-oriented system architecture.


International Journal of Computer Integrated Manufacturing
processes by combining and composing services from
Service Repository/Registry to meet the specific needs
of various requirements. It describes how services can
interact with each other, including the logic and
execution order of the interactions and then constructs
an executable process that may result in a long-lived,
transactional, multistep process plan. With it, the
interactions are always controlled from the perspective
of one of the participants involved in the process. The
Orchestration is the focus of the process management
(PM), workflow, activity monitoring (AM), rules
vendors, etc.
The Governance & Management provides governance and management functionality for the other
components mentioned above. Usually, they include
a trusted and authoritative system of record for
the discovery of services, capabilities for managing
collaboration between technical and business stakeholders in an SOA, tools for enhancing the quality
and conformance of services, capabilities for formalising consumer and provider relationships, enforcing policies within an operational environment to
permit conforming and reject nonconforming service
behaviours at run time and other security problems,
etc.
The Presentation is how services and the applications are displayed to end users. Examples include web
portals, composite application frameworks, mobile
devices, etc.
3.2.

Core functionalities

A practical application of SOA requires an understanding of the complexities and interrelationships of
the SOA concept. Based on the system architecture
proposed above, there are two major functionalities:
governance and management functionality and knowledge-sharing functionality. They are core functionalities that operate together upon the proposed system
architecture to ultimately enable the realisation of
collaborative product development.
3.2.1. Governance and management functionality
The focus of governance and management functionality relies on visibility, trust and control within SOA.
This means creating a foundation that balances the
flexibility and agility with the control and predictability of hardware or software resources. There is
often confusion centred on the relationship between
governance and management. Governance is concerned with decision making, which aims to determine
who has the authority and responsibility for making
decisions and the establishment of guidelines for how
those decisions should be made. Management, on the

117

other hand, is concerned with execution, referring to
the actual process of making, implementing and
measuring the impact of those decisions. Consequently, governance and management work in concert
to ensure a well-balanced and functioning organisation
or inter-related organisations.
There are primary capabilities about governance
and management functionality that include:
(a) Publishing and discovering services so that
services can be found and reused. Standards

based registry that means a simple way
to discover and publish reusable services.
Repository that provides a way to capture,
catalogue and discover all of the metadata,
artifacts and relationships in SOA and
capabilities for rich reporting, impact analysis
and synchronisation with other repositories.
Service catalogue that simplifies the process
of publishing and discovering services with a
straightforward and intuitive application for
providers to publish and consumers to discover services.
(b) Creating policies and validating conformance
so that services put into production are of high
quality. Policy management that takes the time
and complexity out of service validation and
improves the quality and conformance of
reusable services.
(c) Creating consumer and provider contracts to
lessen conflict and set appropriate expectations.
Contract and consumer management that
promotes consumer and provider trust and
conditions that bind service providers and the
consumers who reuse services.
(d) Managing the lifecycle of services to control
services from introduction to retirement.
Lifecycle management that provides control
over versioning and state change of services
from initial introduction to final retirement.
Service level management that manages and
provides visibility into how an SOA is

delivering the actual service in real time and
over time.
(e) Reporting on various dimensions, including
impact analysis to understand how service
changes affect other services and ultimately
business processes. Problem resolution that
facilitates fast problem detection and notification so that despite SOA complexity, performance issues can be diagnosed quickly and
mean time to repair is reduced. Change impact
that decreases the risk of frequent changes in
SOA environments by quickly detecting them
and establishing their impact.


118

Y. Hu et al.

3.2.2. Knowledge-sharing functionality
In this system, capabilities are exposed as standardsbased, shared and reusable services. Services are selfdescribing, independent and loosely coupled, but
controlled through policies, and they can be assembled
ad hoc to orchestrate complex processes, which means
services can be shared among applications despite the
heterogeneous environment.
SOA is based upon the service concept and there
are three main components of SOA standards:
(1) Simple object access protocol (SOAP) provides
the envelope for sending the services messages.
(2) The web services definition language (WSDL)
forms the description for services. A service
provider describes its service using WSDL.

(3) Universal description, discovery and integration (UDDI) registries can be searched to
quickly, easily and dynamically find and use
services.
Based on these standards, there are three roles acting
in SOA; named service provider, service registry and
service consumer. Service provider creates services and
generates service description as service interface
according to WSDL, then registers the services to
service registry using UDDI; service registry maintains
information of services that providers created; and
service consumer (client) finds these services in service
registry using UDDI and interacts with service
provider by SOAP to bind the services for usage.
This is illustrated in Figure 2 below.
On the basis of Figure 2, this system can realise
knowledge-sharing functionality with the mechanism

Figure 2.

Web service sharing mechanism in SOA.

as: based upon the Base Resource, such as database,
knowledge base and expert system, or other legacy
applications, the Integration gives interfaces to connect
and utilise these entities; through the interfaces, service
providers can wrap the capabilities of these entities,
such as domains’ knowledge or other information they
want to share, as services and register the services to
the Service Repository/Registry; furthermore, complex
processes can be composed in the Orchestration by

various services from the Service Repository/Registry;
then clients use the Presentation to find the services
or processes they need, and get the detail service
descriptions to directly communicate with service
providers to acquire the capabilities of the entities,
for example domain knowledge or other information.
Consequently, it makes for true knowledge-sharing
among participants of this service-oriented system.
This research is concerned with the collaborative
product development in which tool and die makers are
involved. Quite typically, the tool and die industry
comprises a complete process chain, starting with tool
and die design and leading to tool and die assembly
and tryout, to act as a connecting link between product
design and serial production. Thereby, collaborative
work among them is needed. Figure 3 describes a
simple example of the collaborative work in tool and
die industry. In Figure 3, the information generated
from the customer requirements is passed to designers.
According to the information, designers construct part
models and then tool and die makers take charge of
tool and die design. Especially, there need to be
evaluation activities to make sure that the part,
designed by designers according to customers requirements, is possible and suitable for future manufacturing by tool and die designed by tool and die makers,


International Journal of Computer Integrated Manufacturing

Figure 3.

119

The collaborative work process in tool and die makers involved development.

and the same to simulation activities that mainly
validate the tool and die model for the future
manufacturing usage, so that collaborations are
required among these activities and knowledge is
shared among the partners. In this research, the need
of capturing and delivering knowledge in the collaborative work is fulfilled by providing specific advice in
accordance with relevant expertise.
Knowledge can be represented in various ways:
rule-based representation, logic, semantic networks,
frames and model-based representation, etc. The rulebased representation consists of conditions evaluating
to true or false and actions to be taken depending on
the results of the conditions, which form rules in a
certain domain. In this way, knowledge is represented
in the form of production rules and rules are written as
IF-THEN statements to provide a formal way of
representing strategies, knowledge and recommendations. This method is the most popular and versatile of
all the representation schemes, and a rule-based
knowledge-representation method was adopted to
develop the knowledge base in this research. As the
research is for tool and die industry, there are
abundant items of rules related to tool and die design
and part manufacturability. For example (Dunbing
Tang 2004):
(1) The diameter d of a hole should be greater than
the part thickness t by a factor k dependent
upon the part thickness, namely, d 4 k*t.

(2) The width of a slot or projection should be
greater than the part thickness by a factor
dependent upon the part thickness.
(3) The notch angles in a profile should always be
rounded.
(4) The inside bending radius of a rib should be
greater than, or equal to, twice the sheet
thickness.
(5) As to the extruded hole with screw threads, the
height of the extrusion is limited to the stock
thickness.
(6) Set-outs should be limited in height to one-half
the stock thickness. If the set-out is made
hollow, it is possible to obtain a height of
approximately 1.5 times that of the stock
thickness.
(7) The distance between the subtractive features such as holes and slots should be a
minimum of two times the stock thickness;
three times is preferable from a die-strength
standpoint.
(8) The minimum distance from the edge of a
subtractive feature (such as a hole) to the
adjacent edge of the blank should be at least the
stock thickness, but preferably 1.5 or 2 times
the stock thickness.
(9) The minimum distance between the lowest edge
of the subtractive hole and the other surface
should be 1.5 times the stock thickness plus the
radius of the bending

120

Y. Hu et al.

Figure 4 is an example rule in the knowledge base. In
the real world, rule sets are always huge and stored
in files, for example, in our research, rules are stored in
.drl files. With object-oriented technology, rules can be
integrated with applications made by object-oriented
languages like Java, which is to customers’ advantages.
Figure 5 shows the whole process to integrate rules
with objects.
Rules stored in .drl files are fed into the object
KnowledgeBuilder, and then a new KnowledgePackage is created by the KnowledgeBuilder. After the
package is built, it is added to a newly created
KnowledgeBase. It can then be used to create KnowledgeSession. Customers use the object KnowledgeSession to fire rules.
To bind the rules of domain knowledge with part
models, an XML-based messaging method is adopted.
Usually, part models are often created by commercial
CAD software in different formats, which make it
difficult to share part models among different systems.
However, in many conventional CAD software, part
models now can be described as well-organised feature
trees that can be subsequently transformed as XML
format files and outputted to external applications.
Thus, XML is used for representing the information of
features of part models. Figure 6 shows an example of
the feature-based XML representations. In Figure 6, a

part model was built in Inventor and then transformed
into a feature-based XML representation. According
to the features containing in the XML representation,
the part model could be rebuilt in NX after the NX
received the XML representation.

XML promotes standardisation of data format and
interoperability among different computing systems
over the Internet. With the standard of interoperability
of XML, the rules of domain knowledge and the
service-oriented mechanism presented above, services
can receive XML representations as input data for the
rules of domain knowledge to judge the design of part
models. The features of part models in the term of
XML representations can be shared and analysed
collaboratively in CPD participants by services
over the Internet. A simple generic example of this
collaborative work in tool and die industry is illustrated
in Figure 7.
This example has trust boundaries between two
organisations; namely, Die Design and Part Model.
It is assumed that these two organisations are peer
entities that have an interest in collaborative work.
Die Design retains the evaluation service, wrapping
rules of domain knowledge about part manufacturability or design optimisation for part model evaluation and exposing it to partner users via service
description as a service interface. Part Model has
a process of design evaluation that is involved in
the collaboration with Die Design, due to Part
Model needs the evaluation knowledge retained by
Die Design. Evaluation service can be invoked by

receiving a request from design evaluation with
the part model in the term of feature-based XML
representations. According to the rules of domain
knowledge, the expert advices about the features
of part models can be generated based on XML
representations and exposed to outside by the service
and then design evaluation can acquire the expert
advices through the service interface from outside
and make any modifications if necessary to optimise
the part model.
4. System implementation
4.1. The implementation details

Figure 4. An example rule created in the knowledge base of
the case study.

Figure 5.

The whole process to integrate rules with objects.

A prototype system has been implemented based
on JBoss SOA platform, and is combined with a
commercial 3D CAD system (Pro/E) and a commercial
database (Microsoft SQL).


International Journal of Computer Integrated Manufacturing

Figure 6.

121

An example of the feature-based XML representation.

The JBoss SOA platform enables enterprises to
integrate services, handle business events, and automate processes more efficiently, linking data, services
and applications across the value chain. In the
prototype system, JBoss SOA platform is the infrastructure that provides the SOA fabric, integration
and management to build SOA. JBoss SOA platform
facilitates participation in collaborative works, and
connects the full value chain.
In the prototype system, a part model is created by
Pro/E. Pro/E is a popular CAD system and can
construct the feature tree of the part model. Based
on Pro/E system, we redeveloped a test function to
transform the feature tree of part model into XML
representations and output the representations as
XML files. These XML files are platform-independent
and can be sent as requests to evaluation services.

The implementation of evaluation services is based
on object-oriented technologies, such as Java language
and we develop service interfaces exposed on the
Internet by service registry, such as UDDI. Through
these interfaces users can invoke evaluation services
and send their feature-based XML representations
as requests and based on the requests, rules in the
knowledge base are called by evaluation services to
fulfill the evaluation tasks.
The knowledge base is implemented by objectoriented language based upon the JBoss Rules

(Drools) that is also a rule engine. However, the
XML document is an ASCII text file, and it can not
be used as an object for Drools. To parse XML
format files for Drools, Java Architecture for XML
Binding (JAXB) was introduced in our research. JAXB
can map Java classes to XML representations, which


122

Y. Hu et al.

provides two main features: the ability to marshal Java
objects into XML and the inverse, i.e. to unmarshal
XML back into Java objects. With JAXB, Drools can
judge if a rule is satisfied or not in the term of featurebased XML representations. Afterwards, based on the
judgments, specific expert advices are generated and
can be sent back to the requester.
To make the illustration clear, we choose a simple
case between one part designer and one die-maker
without complex processes. The next section describes
the prototype system by the case study.

4.2.

A case study

Using an example part, the following case study
illustrates how the part designer cooperates with the
die-maker in the intelligent service-oriented collaborative environment.

The die-maker maintains the knowledge about the
manufacturability of part design, and based on the
prototype system, the die-maker constructs a knowledge base by the rule-based knowledge-representation
method and wraps the knowledge as evaluation

Figure 7.

A simple generic example of collaborative work in tool and die industry.

Figure 8.

A sample interface of the registry facilitator.


International Journal of Computer Integrated Manufacturing

Figure 9.

An example part evaluation.

123


124

Y. Hu et al.

services and then registers them into service registry to
be accessible for external partners. Through the SOA
registry facilitator (its UDDI Browser in JBoss SOA

exampled in Figure 8); the part designer as a customer
can search and find suitable published die-maker’s
evaluation services. After choosing appropriate services, the part designer makes an agreement with the
die-maker about the collaborative evaluation work and
gets the service interface. Then the part designer
transforms the feature tree of part models into XMLbased representation files and sends the XML representations to the die-maker partner via service interface as evaluation requests. After receiving the feature
information and part model from the part designer, the
die-maker parses the XML representations and carries
out the manufacturability evaluation and other evaluation if needed on the design according to the rules
stored in the knowledge base. Any design flaw detected
in the process is notified back to the part designer to
make necessary modification (Figure 9).
5.

Conclusion

In this work, internet-based intelligent service-oriented
system architecture for collaborative product development has been proposed. Identification of the aim and
methodology of this architecture is carried out and
some primary capabilities have been discussed. Based
on features, a case study has been carried out oriented
to part analysis.
The proposed architecture does not aim to replace
existing systems in companies but rather to be an
enabler tool for communicating and sharing knowledge among the geographically distributed CPD
participants. Achieving higher quality, lower cost and
short cycle time is the goal of collaborative product
development. The result shows that service-oriented
architecture platform is considerable promise for
collaborative product development.

The advantages of collaborative product development system architecture proposed in this research are
as follows:
(1) The architecture is based on SOA infrastructure and works in distributed pattern. Through
conventional user interface, remote users can
share design and manufacturing knowledge in
real-time.
(2) The architecture could be used in the early
stage of product design.
(3) Platform independent and heterogeneous
compatible
(4) With the knowledge base, collaborative product development is supported in tool and dies
industry.

For future research activities, emphasis will be put
upon refining the architecture, integrating it with
service-oriented engineering, like automated and dynamic service composition, and constructing autonomous service-based collaborative product development
environment.
Acknowledgement
The authors are grateful to the anonymous referees for
the critical but constructive comments on the original
manuscript.
References
Beiter, K.A. and Ishii, K., 2003. Integrating producibility
and product performance tools within a Web-service
environment. Proceedings of ASME DETC/CIE 2003,
DETC2003/CIE-48237, Chicago, IL, 2003.
Bruce, M., Leverick, F., and Littler, D., 1995. Complexities
of collaborative product development. Technovation, 15
(9), 535–552.
Cassidy, P., 1994. Multimedia comes of age. CIO, 7 (14), 58–

64.
Choo, W.C., Detlor, B., and Turnbull, D., 2000. Web Work:
information seeking and knowledge work on the world wide
web. Dordrecht: Kluwer Academic Publishers.
Chung, M.J., Kwon, P., and Pentland, B., 2000. MIDAS:
A framework for integrated design and manufacturing
process. In: Proceedings of the SPIE the International
Society for Optical Engineering, Vol. 4192, 348–355.
Culley, S., Boston, O., and McMahon, C., 1999. Suppliers in
new product development: their information and integration. Journal of Engineering Design, 10 (1), 59–75.
Dunbing Tang, 2004. An agent-based collaborative design
system to facilitate active die-maker involvement in
stamping part design. Computers in Industry, 54, 253–
271.
Fagerstroem, B. and Jackson, M., 2002. Efficient collaboration between main and sub supplier. Computers in
Industry, 49 (1), 25–35.
Fraser, P., Farrukh, C., and Gregory, M., 2003. Managing
product development collaborations – a process maturity
approach. Journal of Engineering Manufacture: Proceedings of the Institution of Mechanical Engineers, Part B,
217 (11), 1499–1519.
Gero, J.S., 1990. Design prototypes: a knowledge representation schema for design. AI Magazine, 11 (4), 26–36.
Gsell, H., 2006. A content management framework for
product development. In: Fourth Seminar and Workshop
on Engineering Design in Integrated Product Development,
Zielona Gora, Poland.
Handfield, R.B., et al., 1999. Involving suppliers in new
product development. California Management Review, 42
(1), 59.
Huang, G.Q., Huang, J., and Mak, K.L., 2000. Agent-based
workflow management in collaborative product development on the internet. Computer-Aided Design, 32 (2),

133–144.
Johanson, B. and Krus, P. A Web service approach for
model integration in computational design. In: Proceedings of ASME DETC/CIE 2003, QETC2003/CIE-48196,
Chicago, IL, 2003.