This paper aims to develop a system for product modelling in which all of the information generated during a design process can be represented. An integrated product model is an important basis for concurrent engineering by providing a shared representation of the evolving design for team members. This paper focuses on how design specification data is managed in the system.
This paper aims to develop a system for product modelling in which all of the information generated during a design process can be represented. An integrated product model is an important basis for concurrent engineering by providing a shared representation of the evolving design for team members. This paper focuses on how design specification data is managed in the system.
This paper aims to develop a system for product modelling in which all of the information generated during a design process can be represented. An integrated product model is an important basis for concurrent engineering by providing a shared representation of the evolving design for team members. This paper focuses on how design specification data is managed in the system.
Johan Malmqvist and Peter Schachinger Design Specifications, Product Models, Computer-Aided Design, Design Methodology TOWARDS AN IMPLEMENTATION OF THE CHROMOSOME MODEL - FOCUSING THE DESIGN SPECIFICATION 1 Introduction The objective of this work is to develop a system for product modelling in which all of the information generated during a design process can be represented. This information ranges from abstract information like design specifications to detailed information like geometry models. Such a system will have advantages compared to currently available commercial systems by simplifying the reuse and redesign of old design solutions and by facilitating the verification that a product meets its requirements. Moreover, an integrated product model is an important basis for concurrent engineering by providing a shared representation of the evolving design for team members. In earlier work, one of the authors has presented a prototype version of such a system [Malmqvist 1995ab]. The system is theoretically based on the chromosome model [Andreasen 1992]. This paper focuses on how design specification data is managed in the system. The issues covered includes the basic requirements on the design specification, data models for specifi- cation items, and the relationships between the design specification items and other product models such as the function, organ and component structures and behavioural and life-phase system models. A case study of an expansion tank is used to illustrate to use of the system. 2 Modelling the contents of the design specication This section aims at clarifying what information should be included in a design specification and to formulate that information in an Entity-Relationship diagram. Moreover, the potential benefits of using a computerized design specification that is linked to the product model will be highlighted. A design specification states the task for a development project, including requirements, objectives and relevant information. It aims at focusing the design process at identified customer needs while ensuring that no requirement originating from any stakeholder, life-phase or product aspect is forgotten during the development process. Moreover, it provides a basis for analyzing the consequences of modifying requirements. It also serves to create a common understanding of the project goals for the development team and is a instrument for management control of the project progress. The potential benefits of well-managed work with the design specification includes earlier, fewer design changes, shorter development lead times and fewer quality problems. Hence, the design specification is a central reference document for the design team during the design process. The engineering literature provides several procedures for collecting the information to be included in the design specification. (Confer, e.g., [Pugh 1991], [Pahl and Beitz 1996], [Roozenburg and Eekels 1995], [Ulrich and Eppinger 1995]). The QFD methodology [Akao 1992] further provides a matrix-based framework for translating requirements between different stakeholders in the process. The authors generally put forward three kinds of tools for requirement formulation: methods for gathering customer needs, methods for translating customer needs into engineering requirements and checklists that aim at ensuring that no vital requirement is forgotten. Moreover, recommendations on how to state individual requirements are given: Requirements should be given unambiguous, solution-independent formulations and be clearly linked to customer needs. Requirements should further be stated in measurable (or, at least, verifiable) units. It is further useful to distinguish between demands and wishes. It is further important to recognize that the specification is a dynamic entity which changes and expands during the design process. The high-level, form-independent requirements used during conceptual design may need to be concretized to form-dependent ones during the design process in order to be specific enough to set goals for the embodiment design [Otto 1996]. Moreover, new requirements will be added due to the design decisions made, for instance to connect two selected function carriers or to compensate for negative consequences of selecting a particular design solution [Svendsen and Thorp Hansen 1993]. Different variants of the design process may also require different methods for managing the design specification. In original design projects only a few requirements are known at the outset of the design process, and most will emerge during the process as the result of design decisions. This will necessitate frequent updates of the specification if it is to be complete, and some method of distinguishing original and derived requirements, or there is a risk that requirements that were included to compensate for the effects of a certain design decision remain in the specification, although the decision that lead to their inclusion may have been changed. On the other hand, in redesign projects, such as the design of a new car year model, most requirements will be known in advance (and these are numerous). In this case, good requirements management may support the design team in focusing on the requirements that are important, different or new (called deltaSpecs below). Based on the literature cited above, a formal data model for the design specification information may be constructed (fig 1). The model describes the data items of the specification and the relationships between the requirements. The relationships between the design specification and the rest of the product model is elaborated on in section 3. The model shows that a design specification consists of: Metadata. A standard set of metadata attributes, such as identity, product and subsystem name, description, version, status etc, is needed to administratively keep track of the speci- cation. Requirements. The design specication is decomposed into a set of requirements. Referenced documents. The design specication may further refer to a set documents of various kinds, such as old designs, customer surveys, laws and regulations. The individual requirements items are then described by: Metadata. These are similar to those of the design specication. Value. Each requirement will have value, unit and tolerances specied. Requirements class. Examples of requirement classes include performance and geometry. The set of requirements classes used here is adopted from Pugh [1991]. Category. Requirements can be classied as demands (which can be sub-classied into functional requirements and constraints), wishes (or objectives), options or information. Direction. The Direction attribute of a wish category requirement provides further guidance for its value. For instance, it may be desirable to minimize or maximize the value of the requirement, or to hit a specied target value. This attribute may also refer to a value function that species a rating scale for the attribute. Importance. The importance of the attribute may be ranked on a scale. Competition assessment. This attribute keeps track of the competition, as a list of brands with associated values for the requirement. Stakeholder. The stakeholder is the entity who requires the requirement, for instance, the customer or the government. Responsible. The person who is responsible for the satisfaction of the requirement. Origin. The origin of the requirement is documented as original, derived due to a design decision, compensating or interfacing. IsDeltaSpec. This attribute marks a requirement that has changed since the last model of the product. This is useful in redesign situations. Descriptive documents may also be attached to the requirement. See above. Further, we need to be able to register certain relationships between the requirements: ConsistsOf relationships. A requirement may consist of sub-requirements. According to Svendsen and Thorp Hansen [1993] three basic types of decomposition mechanisms for requirements can be identied: A requirement may be decomposed by distribution, such that it is only relevant to some sub-systems, or it may be decomposed by magnitude, such that the value of the system requirement is a sum of the sub-system requirements. Typical examples of the latter are cost and weight requirements. Finally, it may be decomposed by transformation where the property type of a system requirement is transformed into a different type of requirement. This kind of decomposition is dependent on the chosen solution and on the context. TradeOff relationships. In practice, requirements are not independent. Rather, conicts between requirements (corresponding to the House of Quality (HoQ) roof) are likely to exist. For redesigns, these may be known at the outset of the design process. For original designs, trade-offs are likely to emerge during the design process. InuencedBy relationships. Inuences between requirements is here used in a similarly fashion to the translation between customer attributes and engineering characteristics in the HoQ, i.e, as the HoQ matrix area. This model has been implemented in the current version of the system (see section 4). The system can then support a design team in generating and analyzing design specifications. While this is valuable in itself, a more powerful support for the design process is obtained if the design specification can also be effectively linked to the product model. How this can be achieved will be discussed in the following section. 3 Modelling design specications and product models In this section the relationships between the design specification and a product model are analyzed, with the aim of identifying what relationships are important to be able to represent in a design system. The product modelling framework employed in this work is constituted by the chromosome model [Andreasen 1992], which is based on the technical systems theory [Hubka and Eder 1988]. Hubka states that four different types of models are needed to describe a technical system and the transformation process that it affects. These are termed the process, function, organ and component structures, and are said to define the design characteristics of the system. The chromosome expands this theory by adding genetic information, that captures the origin of the design characteristics (hence chromosome). This is achieved by linking the structures with causal relationships: the process determines the functions, the functions are created by the organs, and the organs are materialized by the components. It has further been pointed out [Mortensen and Andreasen 1996] that in order to derive the properties (weight, cost, etc) of a product it is not sufficient to model only the product. We also need to model the various life- phase systems (parts manufacturing, assembly and so on) that the product meets, since many properties (e.g., assembly cost) are relational in the sense that they cannot be determined without knowledge of the life-phase system. The chromosome model offers a general framework for modelling the various aspects of technical systems and is also strongly linked to design methodology. The theory further provides a flexible design process model, by suggesting that the product chromosome (the set of design characteristics) should be seen as a basic map, on which the progress of the design Figure 1: Entity-Relationship diagram for design specication. Design Specification Requirement Identity Name Description IssuedBy ApprovedBy Version Status CreationDate ChangeDate Requirements Class Identity Name Description IssuedBy ApprovedBy Version Status CreationDate ChangeDate Value Category Direction Importance Competitor assessment Stakeholder Responsible Origin Referenced Documents IsDeltaSpec BelongsTo HasMetadata HasMetadata Referenced Documents HasAttributes HasAttributes ConsistsOf ConsistsOf TradeOff InfluencedBy process is chartered. When the design is finalized, all design characteristics will have been established but the sequence of the design activities, which are described as navigations on the map, is not prescribed. We have in earlier work [Malmqvist 1995b] argued that this property and the fact that the causal relationships constitute the genetic information of the system makes the chromosome model well suited as a basis for a design history system. Particularly, the benefits of using an extended function-means tree as a tool for design history representation were put forward in that paper. We are further expanding on that basis here. Having clarified the contents of the design specification and of the product model we may now proceed to discuss the relationships between the specification and the product model. We suggest that at least four principally different relationships can be identified: Measures relationships. The most obvious link between the design specication is when the performance of the design is to be evaluated against a requirement. This type of relation will then link a requirement and a property model that is used to assess the actual value of the property. A property model may in this context be a computational procedure, a experi- mental test or a subjective assessment. The utilization of this kind of relation facilitates checking that a design meets its requirements by comparing the desired and actual states of the properties. Alternatively, a value function could be used to grade the design according to some scale. CorrespondsTo relationships. Some items of the design specication directly corresponds to elements in the product models. For instance, this applies to functions stated in the design specication that have direct correspondents in the function-means tree and the function structure of the product model. This relationship may also link a design specication item to a solution element (means/organ/component) if the design specication states that a particular design solution should be used (carry-over, design re-use). While such requirements are not recommended in the design methodology literature, which states that requirements should be given solution-neutral formulations, they are common in practice. RequirementsOn relationships. In general, a design decision, e.g., the selection of a particular means to solve a certain functional requirement, will generate new requirements to support the means, to connect the means to other means or to compensate for negative effects caused by the means (e.g., noise). It is then important to keep track of this dependency so that if the means is (ex)changed, its generated requirements will also be (ex)changed. The inclusion of RequirementsOn relationships in the model enables the detection of such requirements. IsSolvedBy relationships. IsSolvedBy relationships link a requirement and the means that has been chosen to satisfy the relation. Note that this kind of relation may also be represented via the function-means tree. The resulting product information model is shown in figure 2 as an Entity-relationshipship diagram. Including the design specification data into this model simplifies the verification of requirements and the analysis of the consequences of requirements changes. Moreover, this kind of integrated product model is an important basis for concurrent engineering by providing a shared representation of the evolving design for team members. 4 Implementation In order to verify and experiment with the ideas expressed in the previous sections, a prototype design system has been developed. The aim of this system is to provide an environment for experimentation with design methods based on design theory. While the implementation of this system is not complete, the main framework exists [Malmqvist 1995a]. The emphasis is placed on verification of theories for managing the information created during the design process rather than on creating a working system for synthesis, which would require a larger knowledge base. METIS Software [NCR Norge A/S 1996], a general tool for developing systems for managing complex information structures, was used to implement the system. METIS provides a generic set of system modelling concepts and operations. Object-oriented techniques (type hierarchies) are then used to create objects, relationships and methods tailored to a particular situation. The objects are defined by attributes and methods, and may also include so-called information elements which contain references and parameters to invoke external processes and systems, such as CAD or word processing programs. Information elements can also be scanned pictures or refer to external database files. This enables the construction of very heterogeneous and complex models. System models can be visualized and edited in network, as well as matrix fashion as indicated in fig. 4. A hierarchic visualization is also available. The system has been tested on a number of problems, some originating from industry. The tests have demonstrated the feasibility of the approach, though there is a need for further verification. Figure 2: Entity-Relationship diagram for design specication and product model. Requirement ConsistsOf ConsistsOf TradeOff InfluencedBy ConsistsOf ConsistsOf Process structure ConsistsOf ConsistsOf ConsistsOf Function structure Organ structure Component structure Means HasInfluenceOn I s S o l v e d B y H a s A l t e r n a t i v e R e q u i r e m e n t s O n ProcRelation FuncRelation OrganRelation CompRelation C o n t r i b u t e s T o M a t e r i a l i z e d B y I s S o l v e d B y R e q u i r e m e n t s O n N e e d s E f f e c t s N e e d s P r o c e s s Func Req Objective Constraint Evaluation Property value Test procedure Computational model Subj assessement Uses I s D e t e r m i n e d B y ConsistsOf ConsistsOf B a s e d O n Measures R e q u i r e m e n t s O n I s S o l v e d B y CorrespondsTo C o r r e s p o n d s T o C o r r e s p o n d s T o C o r r e s p o n d s T o PROPERTIES PRODUCT CHROMOSOME DESIGN SPECIFICATION FUNCTION-MEANS TREE Lifephase system LIFE-PHASE SYSTEMS As an example, consider fig 4 which shows a design specification and a product model for an expansion tank, used to maintain a constant water level in an automobile engine cooling system. The figure illustrates a scenario of tracing the relationships between a requirement and the components that realizes the requirement (function). The basic question is: Which components will be affected by a change to a certain requirement? (or vice versa). In this case, the requirement that designer chooses to examine is that the expansion tank system should alert the driver when the water level is lower than a certain level (1). The designer could then follow the CorrespondsTo relation to the function-means tree to learn that the alert function is triggered by an electrical switch and that it also requires a supporting function: transient oscillations in the water level need to be dampened out or else false alerts may result. This is solved by dampening ribs (2). For further insight into the context of the requirement the designer could follow links to the function structure (3) and the organ structure (4). The organ structure is linked to the component structure (5). By this traversal, it is shown that a change to the alert driver requirement may lead to changes to two component (structures), namely the indicator assembly and the expansion tank, since the latter houses the dampening ribs. It is further evident from the figure that overall structure of the models become quite complex, and that navigating in the information may be difficult. However, the ability to visualize system relationships in matrices is very helpful when analyzing system structures. As an example, a matrix depicting the relationships between the organ and component structures is shown. Such a matrix can be used to analyze the amount of function sharing in the system. It also makes it clear that a changes to the expansion tank may affect almost all of the functions in the system, as there is a (almost) one-to-one mapping between the functions and organs. Similar matrices are available in the system for analysis of any kind of relationships in or between system models 5 Related work The framework proposed in this paper is based on the chromosome model developed at TU Denmark by Andreasen [1992] and co-workers. Design specifications in relation to this model have been adressed by Svendsen & Thorp Hansen [1993]. The present paper builds on that work by connecting the design specification to the different parts of the chromosome model This work further contributes with an implementation that can be used as a test-bed for experimen- tation with product models based on the chromosome framework. This is in our opinion crucial as the chromosome model leads to large and complex product models. Such models are difficult to record, keep updated and analyze on paper. An important precondition for testing utilizing the chromosome model in full is therefore that there exists enabling computer support. The current system can also be compared to systems aimed at supporting Systems Engineering, such as RDD-100 [Ascent Logic 1996]. This system also enables traceability from requirements to functions, and from functions to components. However, its terminology is limited from a (mechanical) design methodology viewpoint: RDD-100 does not distinguish between organs (function carriers) and components and there is no correspondent to the function-means tree employed here as a design history representation. We argue that the same argument may be employed for comparisons between the current system and such that are based on QFD - the current system has a richer vocabulary for supporting design work. Finally, the current system is also related to work that has aimed at developing computational support for conceptual design (Tomiyama et al. [1994]; Sharpe [1996]). While these systems provide more powerful support for synthesis than the current, by including design solution libraries, the design specification is not adressed in detail in these works. Rather, the problem is stated as a function (structure). While this may be sufficient for guiding the search for solutions during conceptual design, a more comprehensive approach towards design specifiations is needed if the system shall be used also during the latter design phases 6 Conclusions The inclusion of the design specification in a integrated product model has benefits in terms of increased traceability of design decisions and simplified redesign due to requirements changes. To achieve this, the contents of the design specification and the relationships between the rest of the product model needs to be adressed. The contents of the design specification are well described in the design methodology literature. The relationships, however, are not equally well treated in the literature. In this paper, four relation types are identified and have been implemented in a product modelling system. The paper demonstrates through an example how the proposed product model traces a design decision from a requirement through the function- means tree and further to an individual component. Figure 3: Trace from a requirement to its realizing components. Expansion tank component structure Coolant Pressure cap spatial relation t spatial relation from Expansion tank spatial relation t spatial relation from energy flow to energy flow from spatial relation t spatial relation t Hose spatial relation t spatial relation from spatial relation t material flow to material flow from Hose spatial relation t spatial relation from spatial relation t material flow to material flow from MVD relation from MVD relation from relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from Indicator assembly spatial relation t spatial relation from MVD relation from MVD relation from MVD relation from MVD relation from Design Specification Environment Weight Maintenance Aesthetics, appearance & finish Materials Life in service Quality & reliability Safety Processes Manufacturing facility Packing Shipping Installation Quantity Target product cost Disposal Competition Product life span Market constraints Shelf life (storage) Time-scales Legislative requirements Standards & specifications Documentation Political & social implications Patents, literature & product data Total system performance specification Interfacing & geometric requirements Testing Function structure Organ structure Component structure Function Means Tree Maintain constant water level in cooling system Expansion tank systemsolved by solves Enable coolant refill Maintain constant water level Enable inspection of water level Enable open/closed lid Connect to maximum level area requires is required by requires is required by requires is required by Refill opening - above maximum water level solved by solves requires is required by requires is required by Lockable lid solved by solves Tube connection solved by solves Tube connection CAD model FM Tree RM . Means Requirements FM SS use used by X-axis consist of part of X-axis in Y-axis consist of part of Y-axis in Expansion tank solved by solves Enable input from high pressure section Store pressurized fluid De-air water Enable output to low pressure section Manage internal pressure overflow requires is required by requires is required by requires is required by requires is required by requires is required by Tube connection Pressure tank Slow steady flow Tube connection Pressure valve solved by solves solved by solves solved by solves solved by solves solved by solves Enable off-line inspection Alert driver if low water level detected when driving Transparent wall between mimimum and maximum levels Electrical indicator at minimum level solved by solves solved by solves Connect to system section with minimum water level Change electrical state when low water level detected Stabilize measured water level requires is required by requires is required by requires is required by Req to functions P_0 Preparation process Expansion tank system in prepared state W_0 Working process P_1 Inspect current water level P_2 Open lid P_3 Fill expansion vessel system with water P_4 Close lid Expansion system water level Expansion vessel: State: lid open, non-filled Expansion vessel; State lid open, filled Expansion vessel system Water Expansion vessel; State: lid closed, filled Required water level W_1 Vary water level W_2 Channel water from high-level section W_3 Detect if water level <= minimum W_4 Detect pressure overflow W_5 De-air water W_6 Alert driver W_7 Channel through pressure valve Water level Water level Water level Low water level signal Open pressure valve signal W_8 Store pressurized water and air W_9 Channel water to low-pressure section Water level Expansion vessel system Water level-affecting perturbations Water level Low water level signal to driver Excess water Water to low pressure sections Expansion vessel system OS RM .. Organ Organ OS RM SS use used by Expansion tank component structure Coolant Pressure cap spatial relation t spatial relation from Hose clip Hose clip Expansion tank spatial relation t spatial relation from energy flow to energy flow from spatial relation t spatial relation from spatial relation t spatial relation from Hose spatial relation t spatial relation from spatial relation t spatial relation from material flow to material flow from Hose spatial relation t spatial relation from spatial relation t spatial relation from material flow to material flow from CS RM - Component Component CS SS use used by X-axis consist of part of X-axis in Y-axis consist of part of Y-axis in Customer Ergonomics MVD relation to MVD relation to MVD relation to MVD relation from MVD relation from PS-OS-RM PS-OS-PAT Organs Functions PS-OS-SS use used by X-axis consist of part of X-axis in Y-axis consist of part of Y-axis in Y-axis consist of part of Y-axis in requirement to from requirement from requirement MVD relation to MVD relation from MVD relation from MVD relation from OS-CS-RM + Components Organs OS-CS-SS MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from Indicator assembly spatial relation t spatial relation from MVD relation from MVD relation from X-axis consist of part of X-axis in Y-axis consist of part of Y-axis in Properties Pressure cap opening pressure evaluation Pressure cap opening pressure test MVD relation to MVD relation from MVD relation to MVD relation from Impact resistance test Dynamic pressure test Static resistance to pressure test Burst test Burst evaluation MVD relation to MVD relation from Ageing test Thread strength test Level indicator tests from requirement from requirement from requirement Sealed hole in tank Electrical switch mechanism Ribs solved by solves solved by solves solved by solves MVD relation to MVD relation from MVD relation from Y-axis consist of part of Y-axis in X-axis consist of part of X-axis in use used by MVD relation from Performance Main function The function of the expansion vessel is to work as an expansion volume for the engine cooliing system Refilling function The expansion vessel should be refillable Enable water level inspection Alert driver if water level <= minimum Minimal leakage requirement to requirement to requirement to requirement to Alert driver if low water level detected when driving Electrical indicator at minimum level solved by solves Connect to system section with minimum water level Change electrical state when low water level detected Stabilize measured water level requires is required by requires is required by requires is required by from requirement Sealed hole in tank Electrical switch mechanism Ribs solved by solves solved by solves solved by solves MVD relation from MVD relation from 1 5 3 4 2 Acknowledgements Mitchell Tan Gin Teck is acknowledged for work on a prototype version of the current system [Gin Teck 1995]. Figure 4: Component structure for expansion tank as a network structure , with attached drawing. The component structure shown here is a subset of the total model shown in g 4. Figure 5: Relationship matrix between the organ and component structures. Expansion tank component structure Coolant Pressure cap spatial relation t spatial relation from Hose clip Hose clip Expansion tank spatial relation t spatial relation from energy flow to energy flow from spatial relation t spatial relation from spatial relation t spatial relation from Hose spatial relation t spatial relation from spatial relation t spatial relation from material flow to material flow from Hose spatial relation t spatial relation from spatial relation t spatial relation from material flow to material flow from p X in p Y in MVD relation from MVD relation from relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from MVD relation from Indicator assembly spatial relation t spatial relation from MVD relation from MVD relation from MVD relation from Drawing attached to expansion tank object + Components Organs E x p a n s i o n
t a n k c o m p o n e n t s t r u c t u r e C o o l a n t P r e s s u r e
c a p H o s e
c l i p H o s e
c l i p E x p a n s i o n
t a n k H o s e H o s e I n d i c a t o r
a s s e m b l y Expansion vessel system Refill opening Lockable lid Tube connection Expansion vessel Tube connection Pressure tank Slow steady flow Tube connection Pressure valve Transparent wall Electrical level indicator Sealed hole in tank Damping ribs in tank Indicating mechanism References Akao, Y., Quality Function Deployment. Integrating Customer Requirements into Product Design, Productivity Press, Cambridge, MA, USA, 1990. Andreasen, M. M., Designing on a Designers Workbench (DWB), Proceedings of the 9th WDK Workshop, Rigi, Switzerland, 1992. Ascent Logic Corp., Introduction to RDD-100, version 4.1, San Jos, CA, USA, 1996. Gin Teck, M. T., Computer Tool for Design Specification Modelling, Technical Report, Machine & Vehicle Design, Chalmers University of Technology, Gteborg, Sweden, 1995. Hubka, V. and Eder, W. E., Theory of Technical Systems, Springer-Verlag, Berlin, 1988. Malmqvist, J., A Computer-based Approach Towards Including Design History Information in Product Models and Function-Means Trees, Proceedings of DTM-95, Boston, 1995a, pp 593-602. Malmqvist, J. Improved Function-Means Trees by Inclusion of Design History Information, Proceedings of ICED-95, Praha, Czech Republic, 1995b, pp 1415-1423. Mortensen, N. H., Andreasen, M. M., Designing in Interplay with a Product Model - Explained by Design Units, Proceedings Tools and Methods for Concurrent Engineering, Budapest, 1996. NCR Norge A/S, Metis Users Guide, version 1.9.1, Horten, Norway, 1996. Otto, K. N., Forming Product Design Specifications Proceedings of the 1996 ASME Design Engineering Technical Conferences, Irvine, CA, USA, 1996, Paper No 96-DETC/DTM-1517. Pahl, G. and Beitz, W., Engineering Design, 3rd edition. Springer-Verlag, Berlin, 1995. Pugh, S., Total Design, Addison-Wesley, Wokingham, 1991. Roozenburg, N. F. M. and Eekels, J., Product Design Fundamentals and Methods, Wiley & Sons, Chichester, 1995. Sharpe, J. E. E., Integrated Platform for AI Support of Complex Design - Part I & II, Concurrent Engineering - Research and Applications, Vol. 3, 1995, pp 295-318. Svendsen, K.-H. and Thorp Hansen, C., Decomposition of Mechanical Systems and Breakdown of Specifications, Proceedings of ICED-93, Den Haag, The Netherlands, 1993, pp 119-126. Tomiyama, T., Kiriyama, T. and Umeda, Y., Towards Knowledge Intensive Engineering, Proceedings of the 1994 Lancaster International Conference on Engineering Design, Lancaster, 1994, pp 319-338. Ulrich, K. T. and Eppinger, S. D., Product Design and Development, McGraw-Hill, New York, 1995. Dr Johan Malmqvist Chalmers University of Technology Machine & Vehicle Design S - 412 96 Gteborg Sweden tel: +46 - 31 - 772 1382 fax: +46 - 31 - 772 1375 e-mail: joma@mvd.chalmers.se