Semantic feature modelling
Introduction
Feature modelling is increasingly being used for modelling products. One of its main advantages over conventional geometric modelling is the ability to associate functional and engineering information to shape information in a product model. This can be, for example, the function of some part of the product for the end-user, or information about the way some part of the product is manufactured.
The basic entity in a feature model is the feature, defined as a representation of shape aspects of a product that are mappable to a generic shape and functionally significant for some product life-cycle phase. An essential aspect of a feature is that it has a well-defined meaning, or semantics, for a particular life-cycle activity.
Two important aspects of the above definition are not well covered by most current feature modelling systems. First, feature semantics is poorly defined, limiting the capability of capturing design intent in the model. Second, feature semantics is poorly maintained, permitting previous design intent to be overruled. Such systems are said to lack validity maintenance facilities.
Current feature modelling systems do provide the user with “engineering rich” dialogs aimed at the creation and manipulation of feature instances. In some systems, however, these “features” occur solely at the user interface level, whereas in the product model only the resulting geometry is stored. Such systems are in essence only geometric modellers. Most other feature modelling systems, although they do store information about features in the product model, fail to adequately maintain the meaning of features throughout the modelling process. For example, a modelling operation on one feature may significantly affect the semantics of other features, without the user even being notified by the system, let alone assisted in overcoming this undesirable situation. Assessing the extent to which feature semantics is kept in a model is therefore a crucial issue in feature model validity maintenance.
Fig. 1 illustrates this idea. Assume that the two longer blind holes in the part were positioned relative to the block right-hand face, whereas the rounded pocket was positioned relative to the step side face, as indicated in Fig. 1a. If the width of the step is increased, the rounded pocket overlaps with the two blind holes, “removing” their circular bottom faces from the model boundary (see Fig. 1b). Consequently, the two blind holes now have the shape imprint of through holes. Stated differently, the semantics of the blind holes has been changed. If the shape now produced was indeed desired, it might have been more appropriate not to use blind holes, but through holes instead, attached to the bottom of the rounded pocket and the bottom of the base block.
In addition to lacking feature model validity maintenance, current feature modelling systems also present unexpected results after some modelling operations that are therefore said to have ill-defined semantics. One of the reasons for this shortcoming is that these systems are too tied to methods and techniques of conventional geometric modelling, e.g. they strongly rely on a history-based notion of the modelling process.
Raising the level of assistance provided to the user in maintaining and recovering model validity is essential to bring feature technology to maturity. In this article, we present an alternative way of feature modelling, in which the problems pointed out for current feature modelling systems are overcome. In particular, the semantics of each feature is clearly specified and maintained during the whole modelling process, and the semantics of each modelling operation is well defined. This new approach is therefore called semantic feature modelling.
In Section 2, current approaches to feature modelling are surveyed, and their shortcomings identified. The basic idea of semantic feature modelling is presented in Section 3, and elaborated in subsequent sections: the specification of feature classes in Section 4, the structure and functionality of the feature model in Section 5, and the feature model validity maintenance scheme in Section 6. In Section 7, a few examples taken from a modelling session illustrate the usefulness of this approach. Finally, in Section 8, current feature modelling approaches and semantic feature modelling are compared on their merits.
Section snippets
Current approaches to feature modelling
Almost all current feature modelling systems are parametric, history-based modelling systems, using a boundary representation as main geometric model. The boundary representation can be used for several applications, e.g. process planning for manufacturing. Examples of such systems are the commercial systems Autodesk Mechanical Desktop [1], Pro/Engineer [27], MicroStation Modeller [28] and I-DEAS Master Series [34].
History-based modelling systems are procedural systems which, together with an
What is semantic feature modelling?
Semantic feature modelling is a declarative feature modelling approach. This means that, in contrast to many current approaches, feature specification and model maintenance are clearly separated. All properties of features, including their geometric parameters and validity conditions, are declared by means of constraints. The main advantage of declarative modelling is the freedom in the type of constraints that can be specified, and therefore in the way a model can be edited and maintained.
In
Specification of feature semantics
Feature class specification involves specification of its shape, its validity conditions, and its interface to the feature model, according to the structure depicted in Fig. 9. For all aspects, constraints are used. These feature constraints are members of the feature class, and are therefore instantiated automatically with each new feature instance.
The basis of a feature class is a parameterised shape. For a simple feature, this is a basic shape, e.g. a cylinder for a hole. A basic shape
The semantic feature model
This section describes the semantic feature model, on which the semantic feature modelling approach is based. First, the important notion of dependency between model entities is introduced (Section 5.1). Next, the two levels integrated in the feature model—the Feature Dependency Graph and the Cellular Model—are elaborated (5.2 The Feature Dependency Graph, 5.3 The Cellular Model), and mechanisms for model maintenance are presented for both levels (5.4 Feature Dependency Graph maintenance, 5.5
Feature model validity maintenance
Embedding validity criteria in each feature class, as described in Section 4, enhances the modelling process in the sense that at the creation of a feature instance its semantics matches the specific requirements of its class. However, this might no longer be the case when in any subsequent operation, the shape imprint of the instance would be arbitrarily modified, and therefore the latter should be prevented by the modelling system.
Feature model validity maintenance is the process of
Example modelling session
The usefulness of the validity checking and recovery mechanisms is illustrated in this section with examples taken from a modelling session with the Spiff system. For this, we use a model that is a variant of the part DEMO07, originally from ICEM-CDC, and made available at the NIST Design, Planning and Assembly Repository [30], a large collection of parts from industry and academia.
The user starts the modelling session by opening the model (see Fig. 20). For each subsequent modelling step, the
Conclusions
There are several important characteristics that distinguish the semantic feature modelling approach from current feature modelling approaches. In this section, these approaches are compared on their merits.
The most salient characteristic of semantic feature modelling is that the semantics of all features is well defined and maintained during the whole modelling process. The use of various constraint types for validity conditions in generic feature classes allows specification of many semantic
Acknowledgements
We thank Maurice Dohmen, Winfried van Holland, Erik Jansen, Klaas Jan de Kraker and Alex Noort for their contributions to the research on feature modelling done in our group. We also thank them, and Jiri Kripac, for valuable comments on a preliminary version of the manuscript. Rafael Bidarra's work has been supported by the Praxis XXI Program of the Portuguese Foundation for Scientific and Technological Research (fct).
Rafael Bidarra is a postdoctoral researcher at Delft University of Technology. He graduated in electronics engineering at the University of Coimbra, Portugal, in 1987, and received his PhD degree from Delft University of Technology in 1999. The subject of his PhD thesis was validity maintenance in semantic feature modelling. His main research interests are feature modelling and collaborative modelling.
References (38)
- et al.
Representation and management of feature information in a cellular model
Computer-Aided Design
(1998) - et al.
Feature modelling and conversion—key concepts to concurrent engineering
Computers in Industry
(1993) - et al.
Generic naming in generative, constraint-based design
Computer-Aided Design
(1996) - et al.
On editability of feature-based design
Computer-Aided Design
(1995) - et al.
On user-defined features
Computer-Aided Design
(1998) - et al.
A repository for design, process planning and assembly
Computer-Aided Design
(1997) Issues on feature-based editing and interrogation of solid models
Computers & Graphics
(1990)- et al.
Expert form feature modelling shell
Computer-Aided Design
(1988) - Autodesk. Autodesk Mechanical Desktop 4.0 User's Guide. Autodesk, Inc., San Rafael, CA, USA,...
- Bidarra, R. Validity maintenance in semantic feature modeling. PhD Thesis. Delft University of Technology, The...
Validity maintenance of semantic feature models
History-independent boundary evaluation for feature modeling
Automatic detection of interactions in feature models
Declarative user-defined feature classes
A semantic framework for flexible feature validity specification and assessment
Multiple-view feature modelling and conversion
A proposal for a feature description language
Feature validation in a multiple-view modeling system
Cited by (0)
Rafael Bidarra is a postdoctoral researcher at Delft University of Technology. He graduated in electronics engineering at the University of Coimbra, Portugal, in 1987, and received his PhD degree from Delft University of Technology in 1999. The subject of his PhD thesis was validity maintenance in semantic feature modelling. His main research interests are feature modelling and collaborative modelling.
Willem F. Bronsvoort is an Associate Professor at Delft University of Technology. He received his master's degree in computer science from the University of Groningen in 1978 and his PhD degree from Delft University of Technology in 1990. His main research interests are geometric modelling, including display algorithms, and feature modelling, including feature validity maintenance and multiple-view feature conversion. He has published numerous papers in international journals, books and conference proceedings, is book review editor of Computer-Aided Design, and has served as program co-chair of Solid Modeling'97 and '99 and as a member of several program committees.