 |
Report on the 2nd Workshop on
|
PDTAG-AM
Product Data Technology Advisory Group ESPRIT 9049 |
Björn Höfling, Thorsten Liebig
Otto-von-Guericke-Universität Magdeburg
Institut für Informations- und Kommunikationssysteme
P.O.Box 41 20, D-39016 Magdeburg, Germany
Email: (hoefling,tliebig)@iik.cs.uni-magdeburg.de
This paper gives a report on the 2nd Workshop on Product Knowledge Sharing and Integration (ProKSI-97). The workshop took place at Sophia Antipolis, France on April 17 - 18, 1997, directly after the Product Data Technology Days 1997 held at the same place. A representative number of participants from industry, standardization bodies (especially STEP) and research attended the workshop. The scope of ProKSI-97 was to discuss current trends in product and knowledge modelling, possibilities for sharing and integration of this information, requirements from users and applications, and future developments, e.g. for standardization efforts. In addition to miscellaneous organizational information this report contains a detailed summary of the discussion and the conclusions. Recommendations are given for future activities in this area. Finally are included all papers presented at the workshop and other written contributions made by the participants.
Product knowledge is widely distributed and stored in a variety of ways: on paper, electronically, and in the minds of individuals. How can this knowledge be shared and integrated in order to minimize revision loops and translation costs between product generations and to obtain more flexibility in the use of product knowledge by different kinds of applications? In this workshop, the participants investigated how product data modelling, applications and standardization efforts can profit from experiences in knowledge representation and vice versa. The workshop provided a discussion forum for specialists in product and knowledge modelling to become familiar with current trends in both fields and to explore the benefits from developing joint perspectives towards product knowledge sharing and integration.
The report is structured as follows: Section 2 describes the organization and scope of the workshop and outlines the main topics. It is followed by the workshop programme with its timetable and a list of all written contributions included in this report. Section 5 as the main part of this report contains a detailed summary of the discussion points in the workhop together with the conclusions drawn by the participants. In section 6, recommendations are given to both the communities represented by the participants and to the European Commission. A complete list of all participants of the workshop is attached. Finally all written contributions and papers presented at the workshop are included.
The reported 2nd Workshop on Product Knowledge Sharing and Integration took place at the International Center for Advanced Communication (C.I.C.A.) in Sophia Antipolis, France, on April 17 and 18, 1997, and it was financed by the European Union's Product Data Technology Advisory Group PDTAG (ESPRIT 9049). The organizers of the workshop were Prof. Horst Nowacki from the Technical University of Berlin, Prof. Dietmar Rösner and Björn Höfling from the Otto-von-Guericke University of Magdeburg, Germany.
ProKSI-97 has been the follow-up event of the successful workshop on ``Product Knowledge Sharing for Integrated Enterprises'' held at the conference ``Practical Aspects of Knowledge Management'' in Basel, Switzerland, on October 30 - 31, 1996 (see Deliverable 2.5 of PDTAG-AM). ProKSI-97 was held in conjunction with the conference PDT-Days 97 at the same place on prior days (April, 15 - 16, 1997).
The workshop enabled discussion and interaction between the following communities:
The participants of ProKSI-97 have been chosen mainly on invitation because we decided to limit the number of participants to 25 in order to enable each participant to play an active role in the workshop. Further request for participation could be accepted after initial invitations and aft a public call for participation in the World Wide Web and by e-mail. The organizers succeeded in their effort to assure that members of all relevant groups from the communities mentioned above were represented. In order to encourage a lively discussion, the number of submitted and invited talks presenting a paper was reduced compared to the first ProKSI workshop.
The main topics for discussion were:
| 9:00 h | Opening, Workshop Objectives (H. Nowacki) Short review of the last workshop 'Product Knowledge Sharing for Integrated Enterprises' held in Basel Oct. 1996 (D. Rösner) |
| 9:10 h | SESSION I - OVERVIEW PRESENTATIONS
Julian Fowler/Alan Boyle (speaker), PDT Solutions, UK: |
| 10:20 h | Coffee break |
| 10:20 h | Nicola Guarino, LADSEB-CNR, Italy: 'Conceptual modelling of product knowledge: towards a unified, well-founded methodology' |
| 11:30 h | SESSION II - APPLICATIONS
Peter Dietz et al. (speaker: Steffen Penschke), TU Clausthal, Germany: |
| 12:30 h | Lunch Break
Hans Grabowski et al. (speaker: Karl Hain), Univ. of Karlsruhe, Germany: Dietmar Rösner et al, Univ. of Magdeburg, Germany: |
| 16:00 h | SESSION III - WORKING SESSION including further short presentations from: M. West, Shell, UK: H. Akkermans, Univ. of Twente, The Netherlands: H. Nowacki, TU Berlin, Germany: S. Kneebone, Univ. of Coventry, UK: |
| 17:30 h | End of Session III |
| 9:00 h | cont. SESSION III - WORKING SESSION
C. Partridge, REV-Eng Cons, UK: |
| 10:15 h | Coffee break |
| 10:30 h | SESSION IV - CONCLUSIONS
Remaining open issues
|
| 12:30 h | Closing remarks by the organizers |
The following papers/talks have been presented at the ProKSI-97 workshop (the speakers are underlined):
| (A.1) | Julian Fowler and Alan Boyle, PDT Solutions, UK: 'From product to product knowledge modelling in STEP' |
| (-) | Nicola Guarino, LADSEB-CNR, Italy: 'Conceptual modelling of product knowledge: towards a unified, well-founded methodology' |
| (A.2) | Peter Dietz, Steffen Penschke, Andreas Ort, TU Clausthal, Germany: 'Strategies for product knowledge management and feedback to design - application examples' |
| (A.3) | Karl Hain, Hans Grabowski, Stefan Rude, Chenguang Liu, Univ.
of Karlsruhe, Germany: Supporting the search for design solutions based on information recognition and automated classification' |
| (A.4) | Dietmar Rösner, Björn Höfling, Thorsten Liebig,
Univ. of Magdeburg, Germany: 'Modelling of product knowledge in the framework of multilingual technical documentation' |
The complete texts of the papers can be found at the end of the report. Unfortunately, Nicola Guarino was not able to deliver the written version of his invited talk until the completion of this report.
The following shorter written contributions to the workshop are included:
| (B.1) | Horst Nowacki: Semantic Precision and Logic in STEP |
| (B.2) | Nicola Guarino: Remarks to the Statement of Horst Nowacki |
| (B.3) | Horst Nowacki: Response to the comments on the semantic precision needed in STEP from Nicola Guarino |
| (B.4) | Chris Partridge: A Business Object Ontology |
| (B.5) | Contributions by Hans Akkermans:
|
| (B.6) | S. Kneebone, G.N. Blount: A Short Term Strategy for Product Knowledge Sharing and Re-Use |
A major need for future discussions was identified at a very early stage of the presentations during this workshop: The need of an interpretation framework for all groups interested in product knowledge sharing and reuse (like industry, research in artificial intelligence (AI), standardization bodies, etc). It should clarify the meaning of terms frequently used in this field in order to facilitate the communication between so different groups. Beginning with the term product and product knowledge, there was a general agreement among the participants that this includes every artefact which can be ``produced'' in an enterprise, not only the material ones. Examples for other types of products are software or services. Another proposals was made to extend our domain of interest towards industrial knowledge which encompasses more than product knowledge, because it includes additional aspects that are not directly linked to a product like processes or the organization of a company.
Among many other notions that have to be clarified we only want to mention two at this point: The term ontology is often used in the area of AI as a means for enabling sharing and reuse of knowledge. What is meant by it in the case of product knowledge and what the benefits of using ontologies are should be made clear in the interpretation framework. Participants working at industrial solutions pointed out not to use scientific terminology when talking to engineers (citation: `If we say to the industry, you need an ontology, they would show us the door').
To represent knowledge about a certain area of interest, one has to describe it at different levels of detail and has to introduce abstraction barriers in order to make it reusable. Distinctions like those into generic, domain-specific or application-specific models are often very different. Especially the scope of the generic model should be agreed upon to form a common basis for the discussion. To give an example, within STEP (Standard for the Exchange of Product Model Data, ISO 10303), the generic model is composed of the Integrated Resources (IR). But there are many other approaches.
A significant portion of the discussion was spent on the question 'How to characterize product knowledge and how should it be represented?' This question emeerged as the major point for building product knowledge for sharing and integration. The problem is that most of the current modelling formalisms for product data which are used in industry, like STEP, have limitations in their expressiveness. This was considered by the participants to be a very fundamental lack, although the formalisms are doing well in enabling the standardized exchange of product knowledge. The two major points of criticism were the inability to express more complex information and the lack of precise semantics.
EXPRESS, the data specification language of STEP, can be used to define data structures and constraints. Its expressiveness can be compared with entity-relationship models. In addition it is possible to define constraints to assure consistency. However, more complex information cannot be described. To illustrate this point: It is not possible to represent the fact that if a complex design construct has a certain number of properties (which may be derived from its parts), it will behave in a certain way. Such additional representation capabilities would considerably enlarge the possible uses of product knowledge. AI can offer representation and reasoning formalisms which might serve as a stimulus for improvements in the expressive power.
A second major point of criticism was the lack of a precise semantics of the concepts used in STEP. At the moment the meaning of an entity is at best described by a natural language statement. There are no restrictions within the standard on how this has to be done and how the different definitions rely on each other. Most of the interpretation is left to the user. This has the consequence that the same entity can be used in different contexts and with different meanings which can obviously create problems when information is shared and reused. One proposed solution to this problem is to build a logical model (or reference ontology) for STEP in order to clarify the meaning of each entity.
It was also discussed how this logical model can be obtained and which additional services could be gained by this approach. As it seems not possible to cite every necessary and sufficient property of an entity, it was suggested to define concepts with the help of some carefully chosen and necessary properties which is often called an axiomatic definition. This will help to add more semantic value to the product models, which is the key for the realization of new services based on product knowledge. The need for such services in future models was a common consensus at this workshop.
It was claimed for example to integrate automatic inferencing services (e.g. consistency and incoherence checking) into the model. Another important service would be to make implicit knowledge explicit to the user. These inference services require methods currently in focus in the field of AI. A move towards a more knowledge based approach will benefit from solutions and methods of the Knowledge Representation/Artificial Intelligence area. This includes a clear-cut distinction between object level and meta level. By means of metaclasses (classes of classes) the behaviour of entire classes (which group instances with similar properties) can be modified, which is currently not possible in EXPRESS.
The clarification of the semantics is especially required for the IR of STEP, because they are reused in every Application Protocol (AP). At the moment, different APs are using the IR in quite a different manner, due to the unclear meaning of these resources. This aspect also contrasts with the view mentioned above to consider the IR as a generic model. The STEP experts among the workshop participants considered the design of the IR a more pragmatic decision in order to use the defined concepts as building blocks. It was also noted that there are great problems concerning the scope of the APs and their interrelationships. Often aspects have to be modelled in an AP because they are not available elsewhere, yet they are so general that they should be shared at a more generic level. Since adding them to the IR takes a longer time of standardization they are sometimes added to the AP. This may lead to inconsistencies when using several AP's simultaneously.
In order to attract more attention not only within the STEP community but also on the outside, from potential users it was suggested to demonstrate the success of future approaches by a feasible test application as soon as possible. This means to show in a nutshell what can be really done in contrast to currently used applications. Therefore, it is necessary to identify example scenarios of design and production applications first. These applications would be strongly connected with the building of concrete ontologies. This requires the clarity of the basic choices of the methodology and discussing pros and cons of different modelling approaches. Participants suggested to classify the functional differences of different methodologies and to use the STEP experience to find basic modelling assumptions. In order to identify them, it was claimed that one needs to describe what falls into the scope of STEP and what does not. Valuable scenarios to focus on were listed: Corporate memories (ontologies for product and process models), models for configurable products and parametric design, process models (e.g. a process library). On the practical side, it was strongly suggested to take care of the implementability of realistic solutions and to spend effort on teaching engineers to become knowledge modellers.
In order to meet the requirements of the user participants argued argued in favour of collecting information for a document about the state of the art. Questions concerning the current use of STEP were: Where and how is STEP used - if STEP is used in companies, then for which purposes is it employed (data exchange or product modelling)? In which way does STEP fullfill the requirements of the applications? What is missing in STEP? The impression of one of the participants was that only a few companies have up to now investigated into STEP. This gave rise to the question, whether the discussed approach is really needed in practice or if it is only an academic playground? (This question was asked more in the sense of a `devils advocat' in order not to miss the requirements.)
Furthermore, it was pointed out, that it is hard to get an overview of the role and hierarchical relationships between existing standards related to product knowledge, even for researchers familiar with the field. One of the reasons for the lack of information on existing standards is the strict information policy of the standardization commissions. The official STEP documentation is fairly expensive and is sometimes not even affordable for some research organizations. All participants agreed that future results and documentation should be published for free (without any fee - like for example for ISO documents). In order not to lose the users, further developements should take care of their requirements and give them easy access to various information documents and examples. This requires the development of browsing tools suitable for uncomplicated navigation through various kinds of documentations.
The relevance of related approaches, research areas and standardization efforts was another item of the discussion. The key questions were: Which ones have to be taken into account, and which fields can give reasonable input to previously mentioned points? The participants especially noted the following points to be relevant:
At the end of the workshop, some technical suggestions for a follow-up stage were collected. It was proposed to build a WWW-page (e.g.\ attached to the PDTAG WWW-page) for having a central point for information. Participants agreed to build a mailing list for future information distribution. In order to ensure future interaction with the STEP community the date of the next STEP meeting in Florence (autumn 97) was announced.
The common major agreement of the participants was the prognosis of the fundamental importance and economical relevance of product knowledge sharing and integration in the future. During the workshop, participants predicted new directions of this field, suggested changes and identified drawbacks of the current state of the art. The following summarizes impressions of the participants regarding the topic of the workshop:
The points mentioned above should be taken into account when deciding on the thematic focus of the Fifth Framework Programme of the European Union.
All participants strongly suggested to continue ProKSI and other workshops in this field. Beyond that, it was proposed to initiate a thematic network. It was also suggested to attract more attention in the STEP community by integrating more industrial partners into future events. The composition of participants from research and industry gathered in this workshop was very good. An equivalent distribution should be targeted for future events. As PDTAG will end its work soon, it cannot support additional events. Therefore, a sponsor is required for the funding of any follow-up ProKSI workshop.
| Prof. Dr.-Ing. Horst Nowacki TU Berlin ISM Sekr. SG 10 Salzufer 17-19 D-10587 Berlin, Germany Phone : +49 30 314 23342 Fax : +49 30 314 26883 nowacki@cadlab.tu-berlin.de Prof. Dr. Dietmar Rösner Mr. Björn Höfling Mr. Thorsten Liebig Prof. Nicola Guarino Mr. Steffen Penschke Mr. Andreas Ort Mr. Karl Hain Mr. Rob Bodington Mr. Alan Boyle Dr. Reiner Reschke Dr. Ing. Arnulf Hagen |
Associate Professor Johan Vesterager Department of Industrial Management and Engineering Building 423 Technical University of Denmark DK-2800 Lyngby, Denmark Phone: +45 4525 4454 Fax: +45 4593 4467 johanv@ipv.dtu.dk Mr. C. J. van der Touw Mr. Jean-Claude Hennet Mrs. Jeanette Breton Prof. Dr. J.M. (Hans) Akkermans Mr. Chris Partridge Prof. A. G. Cohn Mr. Matthew West Dr. Stephen Kneebone Mr. Asko Martio Mr. Hannu Peltonen Mr. Juha Tiihonen |
A.1 |
J. Fowler, A. Boyle:
|
A.2 |
P. Dietz, S. Penschke, A. Ort:
|
A.3 |
H. Grabowski, S. Rude, C. Liu, K. Hain:
|
A.4 |
D. Rösner, B. Höfling, T. Liebig:
|
