A HYBRID DECISION SUPPORT SYSTEM
The joint use of Simulation, Coloured Petri Nets and Expert System
Fabiano A. Hennemann, Ricardo J. Rabelo, José E. R. Cury
DAS - UFSC, P O Box 476, 88040-900 – Florianópolis – SC, Brazil
José V. Canto dos Santos, Arthur T. Gómez
PIPCA – UNISINOS, P O Box 275, 93022-000 – São Leopoldo – RS, Brazil
Keywords: Decision Support Systems, Simulation, Coloured Petri Nets, Expert System.
Abstract: This works aim to propose a Hybrid Decision Support System (HDSS), based in Simulation and Coloured
Petri Nets as modelling techniques of manufacture processes, and an Expert System to assist in its use. The
HDSS provides a friendly interface for the user that, after selecting input parameters, gets a series of data
about the manufacturing process that will assist in the evaluation of its performance as answer. To validate
the proposal, some particular scenes have been tested, with the objective of elaborate a set of proposals for
improving the performance of productive systems, evaluating the impacts from the change on model
parameters and providing a better understanding about the systems considered. The HDSS makes possible
for managers, without knowledge of modelling techniques, manipulate data and interact with the models.
The developed prototype is generic for applying on general manufacturing processes, making it possible to
use it for any industrial plant, since that the input parameters of the model are adequately fitted.
1 INTRODUCTION
The companies must constantly improve its
manufacture processes and its methodologies of
work. For this, becomes necessary the improvement
of the productive process, looking for the reduction
of lead times, costs of production, improvement of
the quality, among others. However, this objective is
very difficult to reach. One of the causes is the lack
of good computational systems that assist managers
in the evaluation of the company and the posterior
decisions. The existence of a tool to support decision
that interacts with a model of the manufacture
process could benefit these companies to analyze the
performance of its processes, to establish schedules
of execution with precision, to relate the operations
and to plan the resources necessary in the
manufacture process of each product type. In
literature, the development of systems based on
some specific approaches is found, with its results
presenting, of course, inherent limitations to the used
techniques. This work presents a Hybrid Decision
Support System that combines the potentialities of
some excellent approaches, aiming at to improve the
quality of the diagnosis and the decisions to be done.
The techniques integrated in this work are
Simulation, Coloured Petri Nets and Expert
Systems.
In this work the shoes matrices production
segm
ent was chosen to evaluate and to validate the
proposal. Its processes can be considered as flow
shop type (Askin & Standridge, 1993), with great
variability and average production, as described in
(Groover, 2001). The research aims, in short, to
study the productive processes in manufacture
environments, with the objective to supply
theoretical subsidies through tools and
documentation, in order to assist administrative
resolutions. The choice of modelling techniques,
whose contributions interact with the processes
managers, through a HDSS, as suggested in (Piesik
& Weglarz, 1999) and (O'Reilly & Lilegdon, 1999)
is a appropriate strategy in this context. The article is
organized as follows: Section 2 presents the
Identified Problems; Section 3 shows the proposed
model and aspects related to the implementation; in
Section 4 the results and the structural validation
with CPNs are shown; and finally, Section 5
presents the conclusions.
251
A. Hennemann F., J. Rabelo R., E. R. Cury J., V. Canto dos Santos J. and T. Gómez A. (2005).
A HYBRID DECISION SUPPORT SYSTEM - The joint use of Simulation, Coloured Petri Nets and Expert System.
In Proceedings of the Second International Conference on Informatics in Control, Automation and Robotics, pages 251-254
Copyright
c
SciTePress
2 IDENTIFIED PROBLEMS
During the mapping of the process operations and in
interviews made with managers of productive
processes, some problems had been identified. These
problems are listed below.
The products manufacture sequence are
frequently unregistered, making difficult the
visualization, understanding and analysis of the
process.;
There are few estimates of necessary
manufacture times. The stated periods are supplied
to the customer based on the experience of the
managers;
Trustworthy estimative about the use of
available resources does not exist;
Absence of a tool to analyze impacts from
investments in new resources;
Extensive amount of rework.
3 PROPOSED SYSTEM
The structure of the proposed system is based on the
problems that had been highlighted by the managers
and in the possible results that each modelling
technique can produce. The HDSS interacts directly
with the simulation model developed with a
Simulator. The use of CPNs complements the
available data on the manufacture process and
analyzes the properties of this model in structural
terms, validating it. Because the simulator and the
CPN generate an enormous amount of information,
the manager is assisted by the module of
intelligence, represented for an Expert System, who
assists in the data interpretation. Figure 1 shows the
basic architecture of the proposed system. Each part
of the system is described below.
Figure 1: Basic architecture of the proposed system.
3.1 User Interface
This module is divided in two parts called:
Managing Module - Data entry and Managing
Module - Results. Using the User Interface, the
manager selects and modifies the desired
parameters, before executing the Simulation. After,
these parameters are saved on text files in the
Repository of Data and the Simulation begins.
Concluded this stage, the results are shown to the
user and stored again in the Repository. This module
was developed using Delphi as programming
language.
3.2 Expert System
The Expert System is divided in two modules. The
first module assists in the definition of the data that
must be selected in the User Interface - Data Entry
and the second provides the interpretation of the
Results, shown in the Managing Module. This
function helps the manager in the interaction with
the User Interface and interpretation of the Results
that the HDSS presents. The Expert System was
developed in shell Expert Sinta.
3.3 Repository of Data
This module stores all the information supplied in
the User Interface Module, the results reached for
Simulator and CPNs, and all the data used by the
Simulator program. This repository also stores the
set of logs from all evaluated scenes. Beyond
serving for future evaluations, the objective of this
base is to make possible that the different scenes of
developed tests are compared, without necessity of a
new simulation. The files are MS Excel spread
sheets.
3.4 Simulator Module
With this module, through the selection of some
parameters, the manager gets the following results:
amount produced, daily production; WIP; graph with
the state of the entities; graph with utilization and
the state of the resources; graph of use and the state
of the production stages; relation of the resources
(humans and machines) that are being used in more
than 60% of the time; measurers with permanence
time in each sector and documentation with process
data. This module was constructed with the software
Promodel.
ICINCO 2005 - INTELLIGENT CONTROL SYSTEMS AND OPTIMIZATION
252
3.5 CPN Module
The CPN module was developed with top-down
approach (Jensen, 1996). Figure 2 shows a small
example of this net. The application of the CPN is
justified because that the majority of the systems are
complex, making traditional Petri Nets inadequate to
represent a model with large dimensions. The
deadlock property (Dicesare et alli, 1993) of the
physical system is verified. The model was
developed with software CPN Tools.
Figure 2: Part of a CPN model
4 RESULTS
Using the collected data of the current process, tests
with HDSS has been carried through. The results of
these tests for the production of matrices are
presented in Table 1. The space limitation imposes
the results shown of a one type of matrix (PUS).
(The amount of results is extensive and related to
several products.) Considering that the reached
results represents the real behavior of the process,
some pointers have been used to mark the
elaboration of alternative scenes of improvements,
focused in the reduction of production time and
increase of productive capacity.
4.1 Scene 1
In Table 1 is shown the resources with the bigger
percentage of use in the productive plant. Because
the majority of these are human resources, more
people training to execute the corresponding stages
were suggested. For such, the resources indicated in
Table 1 had been duplicates and the new results are
presented in Table 2.
Table 1: Current scene
PUS (Model)
Amount Production 32
PUS/Day 1,15
WIP 83
Busiest Resources P1Mod, P2Mod,P3Mod
Busiest Resources P1Maq,P2Maq,P3Maq,P4Maq
In Operation 59,25%
Blocked 26,60%
Wait for Resources 14,13%
Table 2: Scene 1.
PUS (Model)
Amount Production 50
PUS/Day 1,73
WIP 88
Busiest Resources P1Mod, P2Mod,P3Mod,R1Mod
Busiest Resources R4Mod, P1Maq,P2Maq,P3Maq
Busiest Resources P4Maq,P22Maq
In Operation 60,77%
Blocked 25,08%
Wait for Resources 14,13%
The most significant differences between the
current scene and the scene 1 are listed below:
Increase 56% productivity of matrices PUS;
Machine resources appear with index of use
above 60%;
It was concluded that scene 1 introduces
improvements to the current scene, mainly in
relation to the significant addition of production in
the matrices.
4.2 Scene 2
Aiming improve the data found in scene 1, the
changes shown in Table 3 will be carried through
using the HDSS. These changes are based on the
resources that had presented the biggest percentage
of use. Comparing the results of this scene with the
A HYBRID DECISION SUPPORT SYSTEM - The joint use of Simulation, Coloured Petri Nets and Expert System
253
current scene, in Table 4, the following ones can be
detached as main alterations:
Increase of productivity of more than 100% of
matrices of type PUS;
The time of production product PUS falls
significantly;
The values of products in operation, blocked
and waiting resources had remained essencially the
same.
Table 3: Changes in Scene 1
Previous Resources Modified Resources
1 P1Mod 3 P1Mod
1 P2Mod 3 P2Mod
1 P3Mod 3 P3Mod
1 P1Maq 3 P1Maq
1 P2Maq 3 P2Maq
1 P3Maq 3 P3Maq
1 P4Maq 3 P4Maq
1 P22Maq 2 P22Maq
3 R1Mod 5 R1Mod
1 R4Mod 2 R4Mod
5 CONCLUSION
In this work was presented a hybrid decision support
system, using the main properties of the techniques
Table 4: Scene 2
of Simulation and CPNs, with an Expert System for
aid in its use. With the survey of the necessities and
diagnosis, the construction of a generic model of the
manufacture plant was possible, leading to a higher
reliability and quality in the decisions to be taken.
The used techniques were adequate and had
corresponded to the objectives established. The
Simulation was efficient to improve the performance
of the considered system. The CPNs has been
adequate for structural analysis of the model and, in
the tests presented, shown the structural validity of
the scenes.
REFERENCES
Dicesare, F., Harhalakis, G., Proth, J., Silva, M.,
Vernadat, K., Practice of Petri Nets in Manufacturing.
Chapman e Hall, 1993.
Groover, M. P., Automation Production Systems and
Computer-Integrated Manufacturing. 2 ed., 2001.
Jensen, K., Coloured Petri Nets - Basic Concepts,
Analysis Methods and Pratical Use, vol. 1. 2 ed., 1996.
O'Reilly, J. and Lilegdon, W., “Introduction to
FACTOR/AIM,” Proceedings of the 1999 Winter
Simulation Conference, vol. 1, pp. 201 -- 207,
December 1999.
Piesik, P. and Weglarz, J., “Multicriteria Decision
Support for Flexible Manufacturing Systems Using an
Interactive Method,” 7th IEEE International
Conference on Emerging Technologies and Factory
Automation, vol. 1, pp. 733--734, October 1999.
PUS (Model and Matrix)
Amount Production 71
PUS/Day 2,42
WIP 97
Busiest Resources P1Mod, P2Mod,P3Mod,R1Mod
Busiest Resources R2Mod, R4Mod,P1Maq,P2Maq
Busiest Resources P3Maq,P4Maq,P6Maq,P7Maq
Busiest Resources P8Maq,P9Maq,P10Maq,P13Maq
In Operation 57,90%
Blocked 28,24%
Wait for Resources 13,85%
ICINCO 2005 - INTELLIGENT CONTROL SYSTEMS AND OPTIMIZATION
254