TOWARDS AN EXPERT SYSTEM FOR THE MANUFACTURING
SYSTEM PLANNING OF PRODUCTS WITH GRADED
PROPERTIES
Mariana Reyes-Perez, Jan Broekelmann and Juergen Gausemeier
Heinz Nixdorf Institute, University of Paderborn, Fuerstenallee 11, Paderborn, Germany
Keywords: Expert System, Ontology, Knowledge Base, Graded properties, Manufacturing Process Planning.
Abstract: Ontologies open new ways for representing, sharing and reusing knowledge. This paper is based on the in-
vestigations within the Collaborative Research Centre (CRC) Transregio 30. In the CRC, thermo-
mechanically coupled processes are developed and analyzed. They provide the possibility to produce func-
tional graded components. Functional gradation is the targeted and reproducible adaptation of a material mi-
crostructure with the intention of establishing the macroscopic properties of the component. The objective is
the steady progress of the microstructure´s variation through at least one spatial dimension. To support the
manufacturing system’s planning of products with graded properties, we develop an expert system. An ex-
pert system is a software which emulates the reasoning of an expert. One of the components of our intended
expert system is an ontology. It assists researchers with the use and reuse of the acquired knowledge of the
CRC, as well as with communication between the different projects within the CRC. In this paper we ex-
plain the architecture of the intended expert system as well as its elements, placing special emphasis on the
ontology. Then we explain the ontology´s development based on knowledge extraction and representation.
1 INTRODUCTION
Ontologies open new ways for representing, sharing
and reusing knowledge. The Collaborative Research
Centre (CRC) Transregio 30 investigates structures
with a functional gradation and their manufacturing
processes. “Functional gradation is the targeted and
reproducible adaptation of a material microstruc-
ture with the intention to establish the macroscopic
properties of the component. The objective is the
steady progress of the microstructure´s variation
through at least one spatial dimension” (Reyes-
Perez et. al., 2009).
Graded components require a sophisticated
manufacturing process design which considers the
interaction between the manufacturing processes, the
sequences of the manufacturing steps and the attri-
butes of the part, the appliances to be used, the suit-
able tools, the simulations and all the peripheral ac-
tivities that are involved within the process. Due to
the complexity of these types of components, col-
laboration between experts of different domains is
required.
The CRC consists of research groups from engi-
neering, material sciences, applied mathematics,
production and manufacturing technologies as well
as information technology. It is comprised of diverse
faculties of three Universities: Technische Universi-
taet Dortmund, Universitaet Kassel and Universitaet
Paderborn, in Germany.
The research is divided into the project groups
A,B,C and D. Each project group is divided in sev-
eral projects and each project deals with issues of
functional gradation. Nevertheless, each field has a
different topic and approach. Consequently the rela-
tions between each field are implicit, rather than ex-
plicit.
Typical difficulties within such collaborative
works are the communication, the integration of the
generated information and dilemmas caused by se-
mantic and terminology obstacles. The challenge of
these types of projects increases when not only the
research fields are different, but also the collabo-
ration is geographically distributed.
We decided to create an ontology with different
goals in mind. In the short-term:
To surmount the communication problems;
226
Reyes-Perez M., Broekelmann J. and Gausemeier J. (2009).
TOWARDS AN EXPERT SYSTEM FOR THE MANUFACTURING SYSTEM PLANNING OF PRODUCTS WITH GRADED PROPERTIES.
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 226-232
DOI: 10.5220/0002301502260232
Copyright
c
SciTePress
For the integration of the information acquired
in the CRC;
For modelling the domain knowledge of the
project acquired in the CRC;
In the long-term:
As part of an intended expert system which will
help the developers with the manufacturing
planning of products with graded properties.
An expert system is a software which emu-
lates the reasoning of an expert.
The structure of the paper is as follows: First, we
explain some definitions and basic concepts used for
the development of the work. Then we discuss our
ontology and the structure of the intended expert
system. After that, we give details about the ontol-
ogy development. Finally we describe our future
work.
2 DEFINITIONS
Here we review some definitions and basic concepts
used in the development of the work presented here.
2.1 Manufacturing System Planning
There are different approaches and methods used for
manufacturing system planning. Some of them focus
on the creation and evaluation of technology chains
(Brecher et al., 2005) or provide product-process or
even product-process-resource models in combina-
tion with methods to use them (Feldmann et al.,
2008). Nevertheless, existing approaches do not
provide a solution for the manufacturing system
planning of products with graded properties.
At the Heinz Nixdorf Institute we are currently
developing a specification technique for the concep-
tual design of the product (Gausemeier et al., 2008)
and the production system. In this work, we made an
adaptation of these specifications in order to support
the conceptual design of a manufacturing system for
products with graded properties.
A product can be considered using two ap-
proaches, the product design approach (functional-
ity, what is made for) and the manufacturing system
design (embodiment, what is made from). The
manufacturing system development starts with a
conceptual design phase (Gausemeier et al., 2006).
The result of this phase is the principle solution of
the manufacturing system. The description of the
principle solution of the manufacturing system is di-
vided into aspects. They form a coherent system and
are mapped on the computer by partial models (Fig-
ure 1).
Figure 1: Partial models for the domain-spanning descrip-
tion of the principle solution of manufacturing systems.
Thus, the principle solution consists of a coher-
ent system of partial models:
Manufacturing Requirements: These are the
requirements posed to the production system
(e.g. the dimensions of the working spaces,
speed of operation). These requirements are
derived from the requirements stated on the
principle solution of the product (e.g. toleran-
ces, production volume);
Process Sequence: The description of the
manufacturing operations as a chain of pro-
cesses. Each process is characterised by a ma-
nufacturing function and additional attributes;
Resources: These are all the equipment, tools
and personnel that are required for the execu-
tion of processes (DIN69902, 1987). At least
one resource is allocated to each process of
the partial model “process sequence”. Ne-
vertheless, it is possible that one resource real-
ises more than one process;
Shape: We refer to the shape as workspace, the
required floor space of machines or the ac-tive
areas of handling appliances;
These aspects are made of elements like:
Product Properties: For example weight,
structure, surfaces of the product. They are
used to derive requirements on the applied
manufacturing equipment (e.g. tolerances, sur-
face finish, production volume);
TOWARDS AN EXPERT SYSTEM FOR THE MANUFACTURING SYSTEM PLANNING OF PRODUCTS WITH
GRADED PROPERTIES
227
Process: These are all the activities that are
made within a production system. For exam-
ple: Manufacturing processes, logistic pro-
cesses and assembly processes;
Manufacturing Processes: These are the pro-
cesses required to obtain a first shape or state
from the formless original state to change the
material properties. They are taken from the
standard DIN 8580 (DIN8580, 1974);
Logistic Processes: These processes deal with
the procurement, distribution, maintenance,
and replacement of material and personnel.
For example “transport” or “storage”;
Assembly Processes: These are the processes
that are used to add parts to a product in a se-
quential manner. For example “mounting”;
Parameters: Parameters are used to describe
processes and resources. Process parameters
are specific values required by a process. Re-
sources parameters state the range of values
that the resources are capable to perform;
Material Elements: Include all raw materials,
auxiliary materials, parts from suppliers and
trade goods, as well as raw, unfinished and
finished goods (Gienke & Kämpf, 2006);
Material: The material from which the pro-
duct is made, e.g. Steel;
Products: This class contains the analysed
products within the CRC, e.g. crashbox, a
flanged steel shaft and a door interior trim;
Project Groups: There are four main groups
within the CRC:
o Project A: Process Design
o Project B: Material modelling
/ identification of parameters /
experimental validation
o Project C: Numerical Treat-
ment
o Project D: Product Optimiza-
tion Process
Projects: Each project produces, measures or
simulates different attributes of the demon-
strators. The activities and the attributes that
they handle are classified. The relationships
and interdependencies between the projects
were also considered;
Shape of the Product: The geometry of the
product;
2.2 Expert System
An expert system is a computer program that has en-
coded knowledge and information about a particular
field. This knowledge is represented in a machine
readable format. An expert system emulates the de-
cision making and problem solving capabilities of a
human expert (Nikolopoulos, 1997).
According to Negnevitsky (Negnevitsky, 2005) a
rule-based expert system has 5 components:
Input/output interface: Is the communication in-
terface between user and system (input/out-put
of the information);
Inference engine: Is a software that tries to de-
rive answers from a knowledge base. It ma-
nipulates the knowledge in order to solve
problems;
Knowledge base: Stores the domain knowledge.
The knowledge is represented as a set of rules;
Data base: Stores the facts used to match
against the rules;
Explanation facilities: Enables the user to ask
the software how a particular conclusion is
reached.
2.3 Ontology
In the field of informatics, Sure defined: “An ontol-
ogy is an explicit, formal specification of a shared
conceptualization of a domain of interest (Sure,
2003). ‘Formal’ refers to the fact that the ontology
should be machine readable. ‘Shared’ means that it
is a common characterization of an application area.
‘Domain of interest’ indicates that a specific area is
modelled. In other words, ontologies are used as a
form of knowledge representation about a domain. If
the knowledge is categorized and formalized it is
possible to store it within an ontology.
An ontology has 2 types of relationships: hierar-
chical relationships and semantic relationships. With
the first ones, the elements within the ontology form
tree structures, or hierarchies. With the latter ones, it
is possible to link elements from different hierar-
chies. Semantic relationships could be any expres-
sion. Because of this, an ontology represents any re-
lationship between concepts (Tudorache, 2004). For
example “Element A measures Element B” or “Ele-
ment A defines Element B”. In this way an ontology
produces a clear domain model of the applied termi-
nology. If correctly specified, no ambiguities will
occur (Lewis et al., 2001).
Ontologies offer many applications. They solve
the problems caused by semantic obstacles, such as
those related to the definitions and approaches of the
experts of different fields. They also deal with the
management of the knowledge. Ontologies are a
way to represent the knowledge in a machine-
readable format. This representation allows us to use
the knowledge with different reasoning mechanisms
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
228
and problem-solving methods to generate more
knowledge. In our case, we use the knowledge that
we have of the manufacture of products with graded
properties in combination with the procedural
knowledge of a developer to generate a process
chain to manufacture a product with graded proper-
ties.
2.3.1 Ontologies for Manufacturing Systems
Currently, there are different approaches and meth-
ods for using ontologies in the manufacturing do-
main. Lemaignan (Lemaignan et al., 2006) presents
a proposal for a manufacturing upper ontology that
aims to draft a common semantic net in manufactur-
ing domain. Catalano presents the Product Design
Ontology (Catalano et al., 2008), an ontology for
product design. Other ontologies focus on develop-
ment of reconfigurable modular production systems,
for example, the work of (Lohse et al., 2004), a web-
enabled design of modular assembly systems.
Nevertheless, existing approaches do not provide
a solution for the manufacturing system planning of
products with graded properties. Because of this, we
developed OntoMaDa, which is described in section
3.
2.4 Protégé
Protégé is a free, open source ontology editor and
knowledgebase framework. It lets the user create
any ontology with a graphical user interface.
Protégé represents knowledge in the form of
classes, slots, forms, instances, queries and graphical
tools. The classes are the concepts of our system.
The slots are the properties and characteristics of the
concepts. The slots have a value type (e.g. string, in-
teger), a cardinality (e.g. multiple) and a domain. Fi-
nally the instances are the values of the attributes.
JessTab (Eriksson, 2006) is a plug-in for Protégé
that allows the use of Jess and Protégé together. Jess
(Java Expert System Shell) is a rule engine. With
Jess (Freidman-Hill, 1995) it is possible to build
Java software that has “reasoning” capacity. Jess
uses the supplied knowledge in the form of declara-
tive rules. JessTab enables development of programs
that manage Protégé knowledge bases directly. Pro-
tégé, JessTab and Jess are the core of our expert sys-
tem structure.
3 ONTOLOGY OF
MANUFACTURING DATA
In our sub-project we developed an ontology which
aims to help with the design of the manufacturing
system planning of products with graded properties.
Its name is “OntoMaDa” (Ontology of Manufac-
turing Data).
It was constructed with basic elements which al-
low themselves to be used as construction blocks for
the generation of other solutions if the corre-
sponding rules are followed. It integrates experi-
mental data, conceptual knowledge, empirical mo-
dels and concepts in such a way that the developers
can generate new results. Please refer to (Reyes-
Perez et al., 2009) for more information about this
ontology.
Based on this ontology, two enhancements were
made. First, information about the peripheral activi-
ties of the manufacturing process and information
about the sub-projects research and interaction was
included. The ontology facilitates the data exchange
between the different domain researchers. The items
within the ontology have links between them. Each
link is a relationship (hierarchical or semantic).
These links allow researchers to follow a path in
which the relationships explain how the items are
connected. The researcher can start in any item and
then follow the different paths and “navigate”
through the Transregio project. The projects and
sub-projects within Transregio as well as their work,
are explicitly defined within the ontology.
The second enhancement was in the structure of
the ontology. It was changed to facilitate the manu-
facturing system planning conception. In the follow-
ing paragraphs we explain the concept.
We can see the manufacturing system design as
a “chain of decisions” which starts with the re-
quirements of the product. Each link between the
different items of the ontology, offers different pos-
sibilities. Because of this, the selection of the next
link must be carefully taken. For example, one prod-
uct requirement could be the material (e.g. polypro-
pylene, steel). Each possible material implies a dif-
ferent manufacturing process. A specific example: a
thermo-mechanical process can achieve a tailored
graded hardness in the case of a monomaterial,
whereas a chemical process could be used for a
composite material.
This chain of decisions, which we can see as a
“linear and sequential design”, is what we want to
depict in the ontology. But an ontology allows us to
create relations between the items and as a conse-
quence we create a net with these relations. These
TOWARDS AN EXPERT SYSTEM FOR THE MANUFACTURING SYSTEM PLANNING OF PRODUCTS WITH
GRADED PROPERTIES
229
relations could be dependencies, sequences, restric-
tions or any type of relation. Commonly, the manu-
facturing process design has a sequence and starts
with the definition of the product requirements.
Nevertheless, the ontology doesn´t follow a particu-
lar sequence. It is a net of intertwined hierarchies
and items. Each item was carefully classified in or-
der to allow an intuitive use of the ontology. This al-
lows the developers to start a new design based on
previous information and, what is more important:
the process design is not “linear” any more. It could
start not only with the requirements. It could start
with restrictions, for example: use only the available
machines and processes. In other words, the inten-
tion is to enable the user to start his or her search
from any item of a manufacturing system. Then he
or she can navigate through this net, either forwards
or backwards. In this way, the user can see other
possibilities for manufacturing a product.
4 STRUCTURE OF THE SYSTEM
The structure of the expert system is shown in
Figure 2. Its components are:
Data bank (Information): Here is where the in-
formation about the manufacturing pro-cesses,
resources (e.g. machines) and the pro-ject is
saved. Among other things are pictures,
mathematical models, etc.;
Rules module: It is used to define the rules used
to represent the procedural knowledge. The
rules have the format:
If (condition), Then (conclusion)
Ontology: Here are all the elements and rela-
tionships used to model the domain knowl-
edge;
User and developer interfaces: We are building
a user interface and a developer interface
within Protégé. The first one is where the user
states his or her queries to the system. The lat-
ter one is where the developers control and
modify the contents of the system;
Inference engine: We will use the Jess infer-
ence engine as the inference engine of the ex-
pert system;
Explanation facilities: Enables the user to ask
the software how a particular conclusion is
reached. It is made in Protégé.
Figure 2: Structure of our system.
The knowledge base is the core of the system. It
stores the knowledge of the CRC (factual and heu-
ristic). Its three components are: Data bank, rules
module and the ontology.
The user interface is the input interface of the
system. First, the user writes his or her query. Then,
the interface will invoke the inference engine to ac-
tivate the decision rules, and the procedural knowl-
edge will be activated. If the system finds a possible
solution to the query, the solution will be shown in
the user interface.
The expert system is developed in Java. For the
communication between Jess and Protégé we use the
JessTab plug-in of Protégé.
5 ONTOLOGY DEVELOPMENT
To develop the ontology it was necessary to collect
the necessary knowledge of the domain (knowledge
extraction) and then make the formalization for its
further use (knowledge representation). The formal-
ized knowledge was saved in Protégé.
5.1 Knowledge Extraction
Knowledge extraction is the process of collecting the
necessary knowledge for problem solving in the spe-
cific domain. Here is where the knowledge is gath-
ered and structured or classified in a conceptual
model.
Our intended expert system has 2 types of
knowledge: domain knowledge and procedural
knowledge. The first captures knowledge in a
specific area or domain, in this case, all the elements
that form part of the manufacturing system planning
and relationships between these elements. The latter
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
230
captures the manufacturing system planner’s proce-
dure for planning new manufacturing systems.
The domain knowledge was classified in terms
of a manufacturing system, according to the classifi-
cation described in the section 2.1. On the other
hand, procedural knowledge is extracted in form of
steps.
5.2 Knowledge Representation
Knowledge representation is the formalization of the
extracted and classified knowledge, to make it ma-
chine-readable (Nikolopoulos, 1997).
We used an ontology to model the domain
knowledge and it is planned to translate the steps of
the procedural knowledge into rules and save the
rules in Jess.
5.3 Structure of the Ontology
We use the ontology to model the knowledge that
we have of the manufacture of products with graded
properties. Figure 3 shows some of the hierarchies
contained within the ontology.
Figure 3: Part of the ontology.
We can see that different elements of the ontol-
ogy are matched with lines. Each line represents a
hierarchical relation. This hierarchical relation, in-
trinsically, provides the basis for a logical inference
within the knowledge base. Multiple inheritances is
one kind of inference explicitly represented in the
ontology (Lewis et al., 2001). A relation represents
the dependency between classes in a domain. We
used this characteristic in the development of the
query system of our ontology. One element can have
one or more relations. Some of the relations that we
made in our ontology are:
ProjectGroup_A4-“measures”-Specific_temperature
ProjectGroup_D3-“measures”-Internal_friction
There are different ways to structure the informa-
tion. But the challenge is to do it in a useful way.
The generation of the knowledge depends on the re-
lations. If the relations are excessively abstract there
could be loss of information. If the relations are ex-
cessively concrete the complexity will increase sig-
nificantly and confusion might be expected. Up to
now, the information retrieval is made in two ways:
Query tab in Protégé;
Visualization of the ontology with SHriMP
Simple Hierarchical Multi-Perspective (Jam-
balaya Widget, software within Protégé) and
OntoViz;
6 CONCLUSIONS AND FUTURE
WORK
The focus of this project is to help the developer
with manufacturing planning of products with
graded properties, regarding the interaction between
product and production system. To achieve this goal,
we develop an instrumentarium, which, among other
things, includes an expert system. A key element of
this expert system is the ontology explained here.
Future work includes finishing the user interface
and the developers interface. Another task is to
translate the steps of the procedural knowledge into
rules and save the rules in Jess. Once the rules are
saved we are going to join Jess and the ontology by
means of JessTab.
Up to this point, the expert system will be able to
suggest possible manufacturing processes for the re-
quired products, but to know the exact values of the
machine parameters, the user needs to open the em-
pirical models on Matlab. We are working on an in-
terface between Matlab and Protégé. The idea is that
the user receives the optimal machine parameters to
manufacture his or her product, without opening
Matlab.
At the moment a prototype of the software tool
for the description of functional graded components
is developed and programmed. The tool uses the ex-
ported CAD model of the component and transforms
this model into a voxel model (a voxel is the three-
dimensional equivalent of a pixel), where the user
can save the desired properties values to manufac-
ture. We are working on an interface between this
program and Protégé. This interface will be the final
user interface of our expert system.
TOWARDS AN EXPERT SYSTEM FOR THE MANUFACTURING SYSTEM PLANNING OF PRODUCTS WITH
GRADED PROPERTIES
231
ACKNOWLEDGEMENTS
This paper is based on investigations of the collabo-
rative research centre Transregio30, which is kindly
supported by the “Deutsche Forschungs-
gemeinschaft” (DFG).
REFERENCES
Brecher, C., Klocke, F., Weck, M., Meidlinger, R.,
Wegner, H., 2005. Bewertung von Fertigungsfolgen.
In: Eversheim W (ed.), Schuh G (ed.) Integrierte
Produkt- und Prozessgestaltung – Ergebnisse des
Sonderforschungsbereiches (SFB) 361 der Deutschen
Forschungsgemeinschaft (DFG) an der RWTH
Aachen. Springer, Berlin.
Catalano, C.E., Camossi, E., Ferrandes, R.,·Cheutet, V.,
Sevilmis, N., 2008. A product design ontology for en-
hancing shape processing in design workflows. In
Journal of Intelligent Manufacturing. Springer
Netherlands.
DIN69902 Deutsches Institut für Normung e.V. (1987)
Projektwirtschaft Einsatzmittel Begriffe. Beuth Verlag,
Berlin.
DIN8580 Deutsches Institut für Normung e.V. (1974)
Fertigungsverfahren, Einteilung. Beuth Verlag, Berlin
Eriksson, H., 2006. JessTab. Available at:
http://www.ida.liu.se/~her/JessTab/ [Accessed March
2009].
Feldmann, K., Schmuck, T., Brossog, M., Dreyer, J.,
2008. Beschreibungsmodell zur Planung von
Produktionssystemen. Wt Werkstattstechnik online 98-
2008: 156 – 162.
Friedman-Hill, E., 1995. Jess the rule engine for the Java
platform. Available at: http://www.jessrules.com/ [Ac-
cessed March 2009].
Gienke, H., Kämpf, R., 2006. Praxishandbuch Produktion
– Innovatives Produktionsmanagement: Organisation,
Konzepte, Controlling. Carl Hanser Verlag, München.
Gausemeier, J., Frank, U., Donoth, J., Kahl, S., 2008.
Spezifikationstechnik zur Beschreibung der
Prinziplösung selbstoptimierender Systeme des
Maschinenbaus (Teil 2). In Konstruktion 9-2008: 91 –
108.
Gausmeier, J., Feldmann, K., 2006. Integrative
Entwicklung räumlicher Baugruppen, Carl Hanser
Verlag. München.
Gausemeier, J., Ebbesmeyer, P., Eckes, R., 2006. Virtual
Production – Computer Model-Based Planning and
Analyzing of Manufacturing Systems. In Reconfigur-
able Manufacturing Systems and Transformable Fac-
tories. Springer Berlin Heidelberg.
Lewis, W. D., Farrar, S., Langendoen D.T., 2001. Build-
ing a Knowledge Base of Morphosyntactic Terminol-
ogy. Proceedings of the IRCS Workshop on Linguistic
Databases, Philadelphia: Institute for Research in
Cognitive Science, University of Pennsylvania.
Lemaignan, S., Siadat, A., Dantan, J.Y., Semenenko, A.,
2006. MASON: A proposal for an ontology of manu-
facturing domain. In Proceedings of the IEEE Work-
shop on Distributed Intelligent Systems: Collective In-
telligence and Its Applications (DIS’06)
Lohse, N., Ratchev, S., Valtchanov, G., 2004. Towards
web-enabled design of modular assembly systems. In
Assem Autom. Volume 24, Number 3, pp 270-279.
Michels, J.S., 2006. Integrative Spezifikation von Produkt-
und Produktionssystemkonzeption. Dissertation,
Fakultät für Maschinenbau, Universität Paderborn,
HNI-Verlagsschriftreihe, Band 196, Paderborn.
Negnevitsky, M., 2005. Artificial Intelligence: A Guide to
Intelligent Systems. Addison-Wesley, Great Britain,
2
nd
edition.
Nikolopoulos, C., 1997. Expert Systems: Introduction to
First and Second Generation and Hybrid Knowledge
Based. CRC Press.
Protégé Software. http://protege.stanford.edu/
Reyes-Perez, M., Gausemeier, J., Nordsiek D., 2009. On-
tology Development for a Manufacturing Data Base
for Products with Graded Properties. In International
Conference on Information, Process, and Knowledge
Management, 2009. eKNOW '09, pp. 105-109
Sure, Y., 2003. KTweb Knowledge Technology Fact Sheet
– Ontologies. Karlsruhe. Available at: www.aifb.uni-
karlsruhe.de/WBS/ysu/publications/ontologies.pdf
[Accessed March 2009].
Transregional Collaborative Research Centre SFB/ TR
TRR 30. http://www.transregio-30.com
Tudorache, T., 2004. Representation and Management of
Reified Relationships in Protégé, Protégé Conference.
Bethesda Maryland.
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
232