DISCUSSION PAPER

A Short Term Strategy for Product Knowledge Sharing and Re-Use

Engineering Design, Knowledge Sharing and Re-Use, Product Modelling Standards.

1 Introduction

Among the host of new technologies offering support to the engineer, Knowledge Based Engineering (KBE) represents potentially the most significant product development tool to date. Knowledge Based Engineering is a computer system that enables the creation of a fully engineered, best practice design by storing the experience, geometry and data that relates to a product. The ability to re-use product knowledge has enormous commercial implications. Large users will no longer be vulnerable in storing corporate knowledge assets using proprietary software over which they have little or no control.

This paper describes a programme of applied research that has been undertaken to establish the feasibility of integrating Knowledge Based Engineering models with standard product modelling practices using STEP and EXPRESS.

2 A Vision of Knowledge Sharing and Re-use

For almost every new application that is developed a new knowledge base must be constructed. To develop large knowledge bases the concepts of knowledge sharing and re-use must be realised. If this concept is realised the process of building a Knowledge Based System will start by assembling reusable Knowledge Modules. Much of the development effort would then be channelled into creating specialised knowledge and reasoners that are specific to the task of the system under construction. Within this environment existing knowledge bases could be added to shared repositories that will, over time, evolve into large rich knowledge bases, analogous to today's databases. Four main barriers presently impede the realisation of the overall concept of knowledge re-use and sharing [Neches, 1991];

3 Requirements of Knowledge Based Engineering Systems for Product Knowledge Sharing and Re-use

3.1 Current Commercial Motivations

Each commercial KBES has its own development language and therefore representation of design and manufacturing knowledge that, although may facilitate specific functionality, cannot be shared with other proprietary systems. In this respect the core similarities between the application and CAD and KBE tools can be drawn in the motivation to support effective translation between disparate representations. There are three main motivations for the development of a KBES vendor-independent language for representing design and manufacturing knowledge in descending order of priority;

The resources required to develop and implement Knowledge Based Engineering Systems are substantial. However, the implementation of such systems presents a continuous process of maintenance, validation and verification. This continuing process entrenches a company's knowledge assets in the chosen proprietary system or system's representation languages. The longer this process continues the more vulnerable the existing investment is as the Knowledge Based Engineering System vendors' experience their own product life cycle.

The user company demands on the functionality of an existing investment in Knowledge Based Engineering tools will change. This may mean adopting complementary or replacement Knowledge Based Engineering tools. With an inability to transform the existing captured knowledge to another tool's representation this is not a decision which historically has been taken, at the opportunity cost of added Knowledge Based Engineering tool functionality.

3.2 Product Data Interchange Standards

Industry has continually been driven to reduce product time to market and in addressing this issue has, in part, made use of a number of design and manufacturing automation tools such as CAD. Such tools are capable of running on a multitude of hardware platforms under a number of operating systems and interfacing with other proprietary systems. The approach that has been taken to resolve the problem of data exchange is to use industry standards. The concept of data standards or a neutral data file was developed through the development of database interfaces by various CAD vendors wishing to integrate supplier catalogue information.

However, one major barrier to the widespread adoption of this approach is that a widely acceptable data standard is difficult to design, develop and approve such that vendors will invest in adopting the standard by way of providing translators to and from the standard format.

Figure 3.1 An overview of the CAD data exchange initiatives leading to the development of STEP.

The objectives of STEP, as a current standards initiative, primarily within engineering, is to provide a sharable product model. The way in which this has been tackled has been to develop a data definition language, the EXPRESS formalism (ISO 10303). The CAD data exchange initiatives that have preceded STEP are well documented and summerised in figure 3.1.

Figure 3.2 Involvement of users and vendors in STEP and EXPRESS activities (1993) .

Industry at large is supportive of the STEP initiative but is yet to translate this enthusiasm into mass operational implementation of the standard. This is not to suggest that STEP is not being adopted, in contrast large CAE software vendors are now claiming to be STEP compliant. A recent survey of major commercial users and vendors declaring an interest in STEP and EXPRESS activities illustrates this point quite clearly [Davies, 1993]. It was established that although much effort is being channelled into development and basic awareness, only 5% of those surveyed were actively involved in partial implementation of the standards as illustrated in figure 3.2 [Davies, 1993].

4 Knowledge Sharing and Re-use: A Near Term Strategy

Assuming that the development time line of the activities preceding the development of STEP and EXPRESS can be generalised as a loose model for such large standards initiatives a rough prediction can be made as to the development of knowledge representation standards. Knowledge sharing and re-use initiatives have experienced an initial flurry of disjointed activities although major national standardisation initiatives have recently acted as integrators. This is illustrated in figure 4.1 which positions knowledge sharing and re-use initiatives, at different levels, over the historical profile of the development of CAD data exchange standards.

Figure 4.1 An overview of the development of significant knowledge sharing and re-use initiatives.

Using the status of STEP and EXPRESS activities as a baseline for standards recognition, where current knowledge sharing and re-use initiatives are beginning to converge, the horizon for the emergence of standards for knowledge sharing and re-use is probably ten years away. To propose a near term strategy to satisfy these requirements a number of factors must be addressed such as;

In order to protect an existing investment in capturing and formalising corporate knowledge there has to be a means of representing that knowledge in a way that can be implemented in systems which will have differing implementation languages and capabilities. At a fundamental level the development of a vendor independent knowledge interchange format would provide this functionality of which KIF and the Arden Syntax are current examples.

4.1 A Logical Development Path

The only major activities which have co-operated, suggesting a dependency across the knowledge levels are KIF, Ontlingua and Generic-KB. Within the DARPA Knowledge Sharing Initiative the Knowledge Interchange Format (KIF) has been used as a base for the development of Ontolingua, a specialisation of Ontolingua to a frame based knowledge representation scheme then facilitated the development of Generic-KB.

If there is a logical development path to be followed it is suggested that it might be the development of formalisms from which knowledge interchange will result: ontologies, knowledge representations, Task and Problem Solving Models and Knowledge Access facilities.

5 Requirements of a Standard Interchange Language for Knowledge Based Engineering Systems

In order to establish the functional requirements of a knowledge interchange format, those requirements derived from ICAD and Concept Modeller, as representative commercial KBE systems, can be considered as being representative of the commercial Knowledge Based Engineering user community at large.

The task of defining a standard language in which design and manufacturing knowledge may be represented is monumental. However, at a high level the requirements of such a language may be broken down into several subsets including;

Most Knowledge Based Engineering applications will have knowledge within each of these categories, the relative amounts of which depend on the nature of the application. This characteristic has been used in the evaluation of the potential of individual Knowledge Based Engineering applications. In this instance the first five knowledge types have been considered in the initial development of a standard knowledge representation scheme for Knowledge Based Engineering systems.

6 Constructing Test Application Modules in EXPRESS

Test modules were constructed from an operational application developed in Concept Modeller that supports the design of Auxiliary Power Generators. In constructing the test application modules in EXPRESS a structured development process was followed. In the first instance the general sets of Entities were established those being the Main Exciter Assembly, Main Exciter Stator, Winding Overhang, Field Coil Stator, Stator, Field Coil and Coil. The remaining processes were in effect redundant as much of them had been accomplished in developing the Knowledge Based Engineering application.

To then construct EXPRESS models a subclass and super class structure was constructed paying particular attention to the abstract data types of some of the test modules. Using the subclass / super class inheritance tree a logical ordering of the development process was evident working from the most abstract components to the physical components and then final generator assembly.

(define-part field-coil-stator
 :inherit-from (stator field-coil)        
 :properties (           
     height (+ !winding-core-length
     (the non-connection-end height)
     (the connection-end height))
     winding-od (+ !stator-bore-dia           
     (* 2.0 !coil-thickness))                 
     winding-core-length !stator-core-length

     material 'na           
     mass  (+ (the stator-core mass) 
              (the winding-core mass))
                      )

 :subparts ((stator-core 
             :type 'core
           :orientation ((move the :bottom flush
                           with the :bottom of 
                          (the winding-core)))

     height^stator-core-length        
     od^stator-od            
     thickness (/ (- ^stator-od ^stator-bore-dia)
                  2.0)    
)))

Entity field-coil-stator 
SUBTYPE OF (stator, field-coil) 
DERIVE 
height := winding-core-length +
          non-connection-end height +
          connection-end height 
winding-od :=  stator-bore-dia +
               2.0 * coil-thickness 
winding-core-length := stator-core-length 

material := 'na 
mass := stator-core mass +
        winding-core mass  
End Entity  

Entity  stator-core 
SUBTYPE OF (core)               
orientation : string :((move the :bottom flush
                        with the :bottom of
                        (the winding-core)))
DERIVE
height := stator-core-length 
od := stator-od          
thickness := (stator-od - stator-bore-dia) 
             / 2
End Entity

Table 6.1 A comparison of the structure of Concept Modeller code and EXPRESS.

Using the EXPRESS test module 'Field Coil Stator' a comparison between Concept Modeller Code and EXPRESS can be made as illustrated in Table 6.1.

7 Translation of EXPRESS Modules to the Target Knowledge Based Engineering Tool

7.1 Design of the EXPRESS to Concept Modeller Translation Process

The Product Data Exchange using STEP (PDES) activity in the United States has a National PDES Testbed facility at the National Institute of Standards and Technology (NIST) to provide facilities to test and validate the Standard for the Exchange of Product Model Data (STEP). The NIST EXPRESS tool kit is a software library for building software tools for the manipulation of data models written in the EXPRESS language.

The process by which EXPRESS can be translated to the Concept Modeller form is not too dissimilar from the existing mechanisms used by the NIST FED X tool, with a customised output generator. The translation process from the EXPRESS models to the Concept Modeller form were broken down into a number of logical steps those being the translation of Entities, Expressions, Variables and other reference Schemata. These correspond well to the translation requirements of a single Concept Modeller package of Define-Part, rules and mathematical Expressions and variable properties.

8 Development of the EXPRESS to Concept Modeller Translator

In re-using much of the existing FED X mechanisms the majority of the development effort is in the customisation of the existing FED X output module. To customise the existing output module, also using the existing front end structure of the translation process, the FED X working form that the translation process would generate had to be identified and the corresponding operations that had to be performed on it determined. To achieve this in a structured and methodical fashion, lists of the content and type of information held within the current scope of translation were compiled. The processes which would then use this information to generate Concept Modeller code. were described in intermediate Pseudo Code.

To implement each phase of the customised output module the source code of the FED X tool, written in C, was incrementally altered and recompiled. The hardware used for this exercise was a Sun Sparc Work Station with a GNU C compiler to produce an executable FED XK.

9 Testing the EXPRESS to Concept Modeller Translator FED XK

The FED XK translator was used to translate Knowledge Based Engineering modules represented in EXPRESS to the target Concept Modeller code. A number of tests have been undertaken in this manner, then comparisons made between the FED XK output and the original application source code (reflection test).

Reflecting on the requirements of a Knowledge Interchange language, class and attribute definitions have been successfully implemented in EXPRESS with the Sub / Super Type, Entity and Attribute constructs and have been translated into Concept Modeller code.

The ability of EXPRESS to represent product configuration logic as a series of mathematical structures is possible but is not what is required by the 'orientation' statement in Concept Modeller code. This type of statement requires basic modelling of geometric features that might reside in Part 44 of the EXPRESS data model that is dedicated to product configuration. It is therefore proposed that in any further development of this approach, the geometrical representation is separate (in a different Schema) from the description of design intent, communication between the Schema being achieved with the EXPRESS Use and Reference statements. In adopting this approach an advanced 'EXPRESS K' might be developed whilst still benefiting from the STEP focused development effort.

10 Results & Conclusions

The development of standards for knowledge representation has been identified as one of the mechanisms by which knowledge sharing and re-use might be achieved.

The exchange of knowledge modules from one system to another is the most achievable short term mechanism for knowledge re-use to meet the immediate requirements of large investors in the technology.

Existing programming standards and their associated development processes provide an insight into the daunting task of establishing a knowledge representation standard. This has led to the recommendation of the use of existing standards demonstrating a mechanism for knowledge re-use within the context of Knowledge Based Engineering.

EXPRESS as the ISO standard data definition language for STEP has been demonstrated to be a useful foundation for the development of a standard Knowledge Interchange Language for Knowledge Based Engineering systems.

References

Davies M., "Involvement of users and vendors in STEP and EXPRESS”, Doc.RC/93/0028, CADDETC Report commissioned by S.Kneebone, Coventry University, 1993.

Gruber T.R., "Ontolingua: A mechanism to support portable ontologies’”, Working Document Version 3.0, Knowledge Systems Laboratory, Stanford University, 1992.

Lambert S.C., Chappel H.R., "The Benefits of Knowledge Re-use in Knowledge Based Systems: A Case Study”, Proceedings of Expert Systems, 1992.

Neches R., et al, "Enabling technology for knowledge sharing”, AI Magazine, 1991 pp37-56.

E-mail: S.Kneebone @cov.ac.uk