PDTE Heading 1

assistantashamedData Management

Nov 29, 2012 (5 years and 1 month ago)

336 views

Ontology
-
based product data management

Axel Hahn, Fabian Grüning,
Kevin Hausma
n
n, University of Oldenburg, Germany


Axel Hahn spen
t

five years in industry as
a
senior manager in the development of software
systems
for
product data maintenance and exchange
and holds a juniorprofessorship at the
University of Oldenburg now.

1.

Introduction

While modern product data management systems
(PDM
-
systems
)
provide a great variety of
services their main functionalities aim onto creation, storage, and retrieval of product
information. In addition, versioning and access management are
usually

supported as well.
Once
captured
, the data hold within the system
is

reused in multiple ways. Application areas
vary from product development reusing existing parts via production contr
olling to
advertisement issues and customer relationship

and others in the product life cycle
.

Because engineering is the most knowledge intensive process in modern industrial enterprises
knowledge management gains importance to ensure effective developme
nt of innovative
products

[SSR01]
. PDM
-
s
ystems are naturally bounded to take over a main role in handling
explicit knowledge.

In order to deal with these requirements PDM
-
systems are able to
support

different views
showing the same information from differe
nt positions, e.g. stru
ctural views, bills of material

or
assembly progress charts. These views in turn require the data to be stored in a very flexible
and extensible model.

Ontologies are such models

and put special emphasis on knowledge representation.

This is an
issue that is lacking in most contemporary PDM
-
s
ystems.

Many formalization

technologies
were
develo
p
ed but until recently there

was no common standard language. This changed with the
emerging

of Semantic Web technologies
which

provide
formalism
s approximating real world
domains in a both computer and human understandable manner.
The m
ost commo
n standards
are RDF(
S
)

or OWL.

As already stated
PDM
-
systems
available nowadays
d
o not employ
knowledge representations
.

Beside knowledge representation mo
st other (meta
-
)data stored in the PDM
-
s
ystem on macro
level is very similar to ontologies. Therefore the authors propose to apply ontologies as a
general data model for PDM
-
s
ystems
;

t
he intention of this
paper
is to examine

how
PDM
-
systems could

look like

and to investigate pros and cons
.

This paper is organized as follows. Section 2 offers a closer look into
the domain of product
development, it specific needs and challenges

regarding knowledge management
.
In Addition i
t
also summarizes the methodologies

modern PDM
-
systems apply to meet these. Sect
ion 3
introduces ontologies as
a conceptual model for
software systems in general and
for

PDM
-
systems in particular.

Completing
this paper section 4 sketches a

prototype PDM
-
system called
“SmartPDM”, which perfo
rms first step
s

onward to ontology
-
based product data management
.

Section 5 provides a critical conclusion and finalizes this contribution.

2.

Product data
and knowledge
management

P
roduct development is one of the most knowledge intensive processes inside
mo
st companies
[Hahn
03].
C
omputer aided product development today often means concurrent engineering

in
virtual engineering teams
. Multiple developers are employed using
specific

design views and
tools and
provide

important input
for

other lifecycle stages s
uch as testing, fabrication, or
customer relation management.

Nevertheless, all other lifecycle stages are likely to produce
valuable data as well, though scattered and bad
-
formatted sometimes.

PDM
-
systems accompany the whole product lifecycle. This includ
es
and is not limited to
product development
,
production
,
sales/delivery

and
final dispose. All these complex steps and
tasks are likely to be performed by different experts in different places under different focuses.
In addition, they are not stages
in a

predefined

order
,

but

are dynamic and
do

heavily depend
on each others input and results.

Cooperative processes with and intensive information
exchange between the different stages of design imply

the application of knowledge
management technologies to en
sure a competitive product development and product data life
cycle management.
Due to the partial data representation
,

split over engineering tools,
workplaces and
documents
,

information
-

and knowledge
loss appears to be
un
avoidable. Often,
it
is not reall
y lost, but due to the complexity of its storage and representation impossible to
retrieve when needed.

2.1

Knowledge Management in Engineering

Due to its cooperative structure, the complexity of the product, the variety of the required
knowledge along the pro
duct life cycle and reengineering issues knowledge management is one
key to the economical development and production of innovative products

[CCRH00]
.

Due to the fact that product design can be seen as a kn
owledge generation process, most
explicit knowledg
e is stored in the result of the engineering process: the product model.
Because of
its distribution between engineering tools (partial models), workplaces and
sometimes companies and limitations of the product data models significant parts of the implicit

engineering kn
owledge about the interrelation
ships of the partial product models cannot be
made explicit. Examples

are constraints between model entities, design rational
e
s,
requirement
s,

deployment etc. That leads to the paradox that products gain more a
nd more
complexity and the number of partial models rise but model interr
elation
ships
can
not be
modelled in
a sufficient way. Of cause PDM
-
s
ystems address parts of this explicit knowledge
but they are focused on
document

management. Their support of produc
t structures (bill of
material, production structure, functional structure) provide helpful support for managing this
model interrelationship
s

but their structures are often inflexible and their semantic
expressiveness is limited.

2.2

Product Data Management

A
s

core functionality PDM
-
systems
manage the product data in engineering teams
. Product
data mainly resides with
in the specific
document
s determin
ed by the software tools used to
create and
gather this data

(micro models)
.
These include tools for Computer A
ided Design
(CAD) and Engineering (CAE)

or even office
documents
.
In addition PDM
-
s
ystem
s

store
information about these
document
s (meta data or
also known as
macro models) and process
information to support the cooperation in the team.

Micro and macro mode
l build

the product
model
, a

computerized model of the product under development.

The PDM
-
s
ystem has to
maintain the consistency of the product model and support
s

navigation and retrieval in this
model
.

Most s
y
s
tems handle the micro models as binary files
and
are

not
able to
support the
consistency

management for model entities of the micro model.

T
he macro model
is t
he main representation of explic
it engineering knowledge about the
product
beside

the
partial product model stored in
document
s and has

to ref
lect and to support
the company
’s

specific
engineering organis
ation an
d processes. Therefore the PDM
-
s
ystem

and
especially

its macro model structure

has

to be adapted to the individual
requirements.

Virtually every

PDM
-
system available on the

market (such
as Windchill, Matrix One, or Enovia

etc.
)
provide a customization interface to adapt the data model
representing the macro model
s
t
o
red in a data base backend.

Most often the customization process cause a significant part of
the PDM installation budget.

Re
capitulating
, PDM
-
systems need to fulfi
l

demanding needs when trying to close the gap
between sophisticated data represented using different format
s

(product model)

on the one
hand

and
require an extensive customization process

on the other hand
. T
o addres
s these
challenges
a
n ontology
-
based

semantic net

may provide a helpful solution.

3.

Ontologies

Exchanging
and handling
engineering information requires a common understanding o
f
constructs of the model scheme
.
Therefore s
tandards like STEP provide special ef
forts to define
the model semantics. In the Application Protocols the Application Activity Model and the
Application Reference Model are used to describe the semantic of the model constructs from
the user’s

point of view

[ISO94]
.

Despite these efforts the
models still have some degrees of interpretation freedom. In addition,
standards naturally have problems to describe special information entities, which have not been
covered in the standardization process so far.

Generic methods for semantic data descript
ions were developed as a supplement/in addition to
the various standar
ds providing standardized schemes
. The most important progress in this
area can be found in the deve
lopment of the Semantic Web [BF99
] and its related technologies
in the ontology
-
engine
ering [Gru93
] domain.

3.1

Semantic Net

The development of the Semantic Web followed two ideas: Firstly, to invent a high level
navigation and information retrieval method for the Internet
(to chart a map of the internet)
and
secondly to describe information re
sources that can be found
o
n the internet.

In an ontology engineering process the semantic concepts (topics) of a dedicated domain are
defined. This ontology is used to describe and link information resources.

The most important technological base for the

Semantic Web is the Resource Description
Framework (RDF) and RDF S
chema (RDFS) [W3C03
].

RDF is a very easy way to express interrelations between information resources (a semantic
map)

with XML

or to define topics by referencing them to information resourc
es. Furthermore,
RDF can be used in XML
document
s (or HTML) to annotate existing information entities by
adding meta information. A simple example for this is the linkage of a tolerance in a CAD model
with an entry in a requirement management tool. RD
F exp
resses tup
l
e
s

lik
e T(Subject,
Predicate, Object) where t
he object ca
n also be another subject. This is notated like

T(“tolerance1”, “is required by”, “requirement1”) and expressed in XML using RDF

by
:


<?xml version=1.0”?>

<rdf:RDF
xmlns:rdf=”http://www.w3
c.org/1999/02/22
/
rdf
-
syntax
-
ns”


xmlns:s=”http://myschema.de”>

<rdf:Description

about=http://wi
-
ol.de//part/tolerance1>

<s:is_required_by>

requirement1

</s:is_required_by>

</rdf:Description>

</rdf:RDF>


With this technology

the user ca
n define an
o
ntology
fitting his domain
(see next sub
section)
and describe and link the information resources by using semantic specifications thus making
the orientation in the network very easy. The user can navigate through the distributed model
with the help of his

semantic network. The semantic w
eb is a very powerful method to capture
development knowledge and to make it usable for the user and by knowledge engineering tools.
The example
is depicted in f
igure 3.1

where the semantic n
etwork connects the CAD model
el
ements and the requirement definitions

as a semantic navigation map
.


Fig. 3.1

Connecting CAD model entities and requirements

Productm
anager

Top and side connectors

6

Productmanager

Modules for

-

electronic signals

-

energy connectors

-

light signals

-

pneumatics

5

Productmanager

Up to 128 Contacts 1,5 mm²

4

Productmanager

Pricerange

:

Basemodel

80,00



120,00 €

3

Norm ISO ...

Fuse 8A

2

Productmanag
er

Acid environment

1

Origin

Description

#

3D CAD Model

Requirement List

Semantic Map

The m
ap may be used by the engineer to capture implicit knowledge about the product under
development and uses
this information for navigation and reasoning (e.g. to derive depending
entities to identify changes required in other documents during
reengineering
).

3.2

Ontologies

Ontologies describe

the concepts of a specific domain (e.g. product design).
A
n ontology for

the
elements
of this domain
can be defined

by using RDFS and XML for the notation of the tuples
[SEM02]
. Therefore RDFS is capable
of describing cla
sses for subjects, predicates and objects
and to define their relations (e
.
g. inheritance), axioms and cons
traints following the concepts of
computational

logic.

A recent form of o
ntologies using description logic
was

standardized by the W3C as
the
Ontology Web Language (OWL)

[W3C03]. OWL relies on the concepts of description logic
s

[StSt04]
and
provides distin
ct notation for classes (synonymous
with

concepts, i.e. categories
existing in the domain of interest), properties (i.e. relations interconnecting classes)
, instances
of classes and constraints.

It thereby
combines rich semantic expressiveness with high fl
exibility
and computer readability.

An example

for an
ontology is shown by figure 4.
2
.
Details of this
ontology will be described in the next section.

Adding data
,

so called instances
,

generates
the semantic net
as
mentioned in the sub section
above
.

But
ontologies are more than just another way

of storing data. The semantic
expressiveness allows automatic
inferring
. I
llegal states in the
knowledge

base can be detected,
e.g. when the sequence of the construction plan of a product contains an error so that
the
product cannot be assembled in the specified way. Ontologies may even be used to generate
data by themselves so that a construction sequence may be calculated on its own.

These capabilities can be exploited by
interferenzing tools like
reasoners

[StSt0
4]

that work with
the ontology and are compatible because of the
usage

of the standardized format OWL.

To have the maximum value added to the data accumulated in the process of making a product
it is wise to use the ontology to store every data that occurs

in the process. In that way the
ontology gains its maximum expressiveness and can optimally be used for the features
mentioned above. It can additionally be used as a communication platform as the data
contained in the ontology can be reused by every pers
on working at the same or a different
product so that knowledge once it is produced doesn't get lost over time and has
not
to be
produced a second time.

Ontologies can support knowledge intensive business processes as product design by providing
mechanisms

to structure knowledge bases (e.g. catalogues of solution elements), help the
engineer to express implicit knowledge or support communication and interoperability by
providing reusable specification
s

of concepts.

3.3

Appliance of a semantic net

As stated abov
e, m
uch
product
knowledge remains
un
used, because
it exists
implicit
ly

or
remains
chained into a certain
document
. This problem may be resolved by creating a semantic
net interconnecting product related
document
s and their data contents. The semantic net
i
ntroduces an additional intermediate level, helping to bridge the levels of abstraction described
in the previous section. It therefore provides a valuable separation of concerns between data
storage and semantic representation.

PDM
-
systems target very si
m
ilar objectives as ontologies: T
hey structure and manage the
documents during the design process and provide navigation and other services to the product
developer.
It is therefore likely, that the
concept
s

of the semantic net, ontologies and OWL can
be u
sed in the development of PDM
-
systems. This technology comes together with supporting
implementation frameworks like Jena or KAON [Ober04].

They provid
e dynamic extensibility of
the o
ntology during runtime. By using ontologies as a technology to implement
the
(
conceptual
)

macro model of
PDM
-
s
ystem
s

the implementation can benefit from this extensibility. In addition
the description logic and available
interference

tools offer the engineer the conceptual basis to
express knowledge and may support constraints
reasoning or design automatization. In addition
OWL comes with a technology to refer individual model entities of the partial product model
stored in separated documents.

4.

The S
martPDM
Prototype

In the previous chapter we pointed o
ut that the usage of ontol
ogies

for a product data
management
system
provides many advantages
compared to

standard approaches of handling
data. In this chapter we want to introduce a prototype PDM
-
system

that
benefits

of these
advantages, called SmartPDM, which was built at the Car
l von Ossietzky University Oldenburg,
Germany.

A screenshot (see figure 4.
1
) may give an expression of SmartPDM.



Fig. 4.1

Screenshot of prototype SmartPDM

4.2

SmartPDM ontology

SmartPDM consists of four main parts that can be found in many state
-
of
-
the
-
art PDM produ
cts
,
which we call
modules

and make the PDM’s functionality available.

These parts are



product structure management



process management



document management



user management

Basic concepts of the SmartPDM ontology

and their relationship

are shown in f
igure 4.
2
. In

this
paper

we

concentrate on central concepts that satisfy the demands to model PDM
-
specific
information.

The complete ontology is available from the authors and detail
s

the main concepts.


Fig. 4.2


SmartPDM base ontology

The concepts, subconcepts and prope
rties of the ontology provide the needed expressiveness
to model the data and activities of the business processes.

We will provide examples that show the availability of sufficient expressiveness and the way of
using the concepts of the ontology to expre
ss real
-
world information in the context of the
ontol
ogy for every aspect of the PDM
-
system (see above).



A
machine element
(product)

may
have the version 0.8 (version).



Assembling a window (process) may consist of putting together the sheet glass and the
w
indow frame (both tasks)
.



A document (document) may contain the specifications for a nut (product)
.



Stefan

Müller
(user) may be the leader (role) of the project SmartPDM (product)
.

As these examples show the SmartPDM ontology can already be used to express

many facts of
the world of product data management. Furthermore, it can easily be extended to gain an even
higher level of expressiveness.
One

can, for example, add the state where a user is living in to
the user's information by adding the property “stat
e” to the
concept “
user

. From now on the
information of a user also contains the state where she/he lives.

4.3

Application
of the ontology

The ontology is extendable as there can be added further
concepts

those

specialize

existing
modules or provide completel
y new functionality to the ontology. Those new concepts may use
the existing concepts or introduce complete new aspects to the ontology. The process to
provide further functionality is easy to conduct as open standards are used and a lot of
programs are ca
pable to work with the ontology. Furthermore there is no need to change the
existing code of SmartPDM when the ontology gets richer because the needed functionality is
still provided in comparison to standard relational based systems where a change in the
data
format always leads to an
adoption

of the existing code.

As real world objects are represented by an URI (universal resource identifier) every part of a
product (e.g. a bolt

in a CAD
document
, not only the CAD
document

itself) becomes uniquely
identif
iable world

wide. With such a link it is possible to store and share information world wide
to create an integrated development environment which members may be scattered all around
t
he world. A click on such a bolt

in a bill of material may lead to the CA
D
document
, provides
data about the tests that the bol
t

fulfilled and gives the email address of the developer
responsible for the part. As one can see, the ontology collects all knowledge that accumulates
throughout the development process and provides it

in a human unde
rstandable and computer
processable

way.

The concurrent and distributed development gets pushed further by using a web application as
interface to the ontology so that all information can be accessed and entered from everywhere,
anytime.

A
l
l information that the ontology include
s

has to be produced by either the people who work at
the development of the product or semi
-
automatically by the design programs used in that
process.
Additional knowledge may derive from the actual fabrication proce
ss, from testing, or
from customer feedback.
The time
spent gathering and encoding this

information pays of
quickly by providing an integrated knowledge base of all data that is produced in the
development process and the improved communication platform th
at the ontology provides
between the developers. This facility protects the developers
against

duplicate work as every
piece of information is stored, linked and may therefore easily be accessed. The programs that
provide the functionality to extract the m
etadata from the common CAD,
Microsoft E
xcel,
Microsoft W
ord, etc.
document
s that may be stored in the ontology do probably not yet exist or
lack in functionality. But as the software completely relies on open standards it is likely that such
tools will be

available in the near future.

4.4

Smart PDM architecture

SmartPDM uses several techniques that are well established in the area of software
development to guarantee a high quality software product that is easy to maintain and extend.

We will provide a bottom
-
up view of the SmartPDM architecture.

An abstraction layer is used for
data m
anagement

called J
ena

[JENA
]

that

offers
the needed
support for OWL

and has a rich set of methods for manipulating ontologies
.
Using Jena the
ontology itself can be modelled with
any tool that provides consistency to the OWL standard,
e.g. protégé

[PROT], that gives a convenient way of generating and manipulating ontologies in
a graphical environment without having trouble with any notation problem. Jena

uses a
relational database

for persistency

as its backend which was established
by both

MySQL

[
MSQL
]
and

PostgreSQL

[
PSQL
].

T
he four already discussed
modules that provide SmartPDM’s functionality (p
roduct structure
management
,
process management
,
document management
,
user managemen
t
) are built on
top of the data tier. These modules are implemented using
Apache
’s

[APAC]

S
truts

[
STRU
]
framework that provides the structures for the Model
-
View
-
Control paradigm. Using S
truts

model and control are written in J
ava
[
JAVA
] and the view, as Sm
artPDM is a web a
pplication, is
written using JavaServer Page
s

[
JSP
].

Finally
,

the whole application runs in a
T
omcat

[
TOMC
]
servlet container which is
a subproject of
Apache
’s Jakarta

[JAKA]

project
, bringing SmartPDM
on the web.

SmartPDM is written with
extensibility in mind so that it is easy to write new modules for the
application. Those new modules may work on the same data model (i.e. ontology)
and therefore
with the same data
or they may extend the ontology without getting in the way with any existi
ng
module. A fifth standard module, the model manager, exists that takes over such administrative
tasks.

There is a consistency in using open standards like OWL for data modelling and open software
for generating high quality

and

non
-
proprietary
software w
hich components may get
interchanged
or extended easily
due to using open and standardized interfaces

and software
components
.

5.

Conclusions

The devel
opment of the prototype and its application in smaller design
use cases have proven
that the ontology relate
d technologies can be an alternative basis for the data models which can
be found in contemporary PDM
-
s
ystems. Due to its se
mantic expressiveness and extens
ibility
they are a proper method to
support the engineer in capturing

knowledge, which

can not be
ex
pressed with the partial models of the applied engineering tools.

The performance of the
implementation framework proved
to be

sufficient and due to its generic adaptability on changes
of the ontology customization
s

of
SmartPDM can simply be

done by e
ditin
g the used ontology.
Recent

studies under development show that even smaller ontologies support the engineer to
produce explicit knowledge during design.

Of ca
use this academic prototype has

its limitations in functionality compared to
PD
M
-
systems

currentl
y used

in industries. Providing a web based user interface
S
martPDM
support
s

virtual
engineering teams but lacks in replication concepts etc. As in most PDM
-
systems the version
management of
S
martPDM is limited to documents. This implies that there is no v
ersion
management for the macro model
respectively

the semantic net. For ontologies
versioning
technologies are under development. The adoption of these technologies may provide an
additional feature of using ontologies as a data model for PDM
-
systems.

6.

Ref
erences

[APAC]

http://www.apache.org/ (last access August 2005)

[BF99]

Berners
-
Lee, T., Fischetti, M., 1999, Weaving the Web: The Original Design and
Ultimate Destiny of the World Wide Web by Its Inventor. Harper, San Francisco

[CCRH00]

Caldwell, N., Cla
rkson, J, Rodgers, P, Huxor, A., Web
-
Based Knowledge
Management for Distributed Design, IEEE Intelligent Systems, May/June 2000

[Gru93]

Gruber, T.R.: Towards principles for the design of ontologies used for knowledge
sharing. In R. Guarino, N. Poli, editor
, International Workshop on For
-
mal
Ontology, Padova, Italy, 1993.

[Hah
n
03]

Hahn, A., Integration and Knowledge Management Platform for Concurrent
Engineering, Proceeding of the Internationa
l conference on Concurrent
Engi
neering, Espoo, 2003

[ISO94]

ISO 10
303,

Industrial Automation Systems and Integration,
1994,
s.

also

http://
http://www.tc184
-
sc4.org/

(last access June 2005)

[
JAKA]

http://jakarta.apache.org/ (last access August 2005)

[JAVA]

http://java.sun.com/

(last
access August 2005)

[JENA]

http://jena.sourceforge.net/ (last access August 2005)

[JSP]

http://java.sun.com/products/jsp/

(last access August 2005)

[MSQL]

http://www.mysql.com/

(last access August 2005)

[O
ber04]

Oberle, D., Volz, R., Motik, B., Staab, S.: An extensible ontology software envi
-
ronment, in Staab, Studer (Hrsg.), Handbook on Ontologies, pp. 311
-
333, Springer,
Berlin, 2004

[PROT]

http://protege.stanford.edu/ (last access August 2005)

[PSQL]

http://www.postgresql.org/

(last access August 2005)

[SEM02]

Staab, S.; Erdmann, M.; Mädche, A.: Ontologies in RDF(S), ETAI Journal, 6, 2001

[SSR01]

Szykmann, S., Sriram, R., Regli, W., The Role of Knowledge in Next
-
Generation
Product Development Systems, ASME Journal of Computation and Information
Science in Engineering, Vol 1, Num 1, 2001

[STRU]

http://struts.apache.org/ (last access August 2005)

[StSt04]


Staab, S., Studer, R. (Hrsg.): Handbook on Ontologies, Sprin
ger, Berlin, 2004

[TOMC]

http://jakarta.apache.org/tomcat/ (last access August 2005)

[W3C03]

W3C,

“Resource Description Framework (RDF): Concepts and Abstract Syntax”,

2003, s.

http://www.w3.org/TR/rdf
-
con
cepts/

(last access June 2005)

[Weh00]

Wehlitz, P.: Nutzenorientierte Einführung eines Produktdatenmanagementsystem