Geometric and Feature modelling

Validity Maintenance in Semantic Feature Modeling

Summary

Computer-based techniques for supporting product development have evolved rapidly in the last two decades. Mainly driven by market and quality demands, concepts like rapid prototyping and concurrent engineering have been proposed. Such concepts require more, and more complex, information to be stored in so-called product models. For example, information about the shape of a product, traditionally stored and processed using geometric modeling techniques, is now required to be integrated with other types of product information (e.g. function, manufacturing and assembly information). Feature modeling constitutes an important milestone in this evolution.
Feature models can combine shape information and functional information, which makes them a versatile product representation suitable for the integration of many product life-cycle activities (e.g. design, manufacturing planning and assembly planning). For the success of this integration, a key role is played by a well-defined specification of the meaning, or semantics, of features, clearly associating the shape aspects to their desired functionality. Feature-based modeling systems should correctly interpret and maintain those associations in the feature model, throughout the whole modeling process. This is usually called feature model validity maintenance.
Current feature modeling systems, however, are still very much tied to methods and techniques of conventional geometric modeling systems. Among other drawbacks, they offer only restricted facilities for defining feature semantics, and often fail to preserve this semantics as the feature model evolves.
This thesis presents a new feature modeling approach –designated semantic feature modeling–, which overcomes the validity maintenance problems of current feature modeling systems.

In semantic feature modeling, feature specification is done declaratively in feature classes, using a variety of constraint types. A feature class is a structured description of all properties of a given feature type, and includes the validity conditions that all its feature instances should satisfy.
A two-level semantic feature model has been developed to represent a product. The first level –called the Feature Dependency Graph– consists of a set of interrelated feature and constraint instances: the entities on which all modeling operations are performed. The second level contains an evaluated geometric representation of the product in the so-called Cellular Model. Its most important property is that both the generation and the interpretation of the Cellular Model are independent of the chronological order of feature creation in the model. The two levels are integrated in the semantic feature model, which disposes of mechanisms for automatically maintaining the consistency between them. In addition, it supports a variety of queries at both levels, making it possible to perform, among other things, effective model validity maintenance.
Feature interactions, which arise from modeling operations such as the creation of a new feature or the modification of an existing feature, are among the main causes of feature semantics violations. Such phenomena are therefore thoroughly analyzed and classified in this thesis, and a variety of interaction detection algorithms is also presented.
The validity maintenance scheme presented here basically monitors each modeling operation, in order to assess the conformity of all features in the semantic feature model with their validity criteria. This is achieved by maintaining all constraints, using various constraint solving techniques.
Each invalid situation detected is analyzed by a validity recovery mechanism, which gives the user explanations and context-sensitive hints to overcome the situation. The user gets thus valuable assistance in creating valid models only, containing features with well-defined semantics only.

The semantic feature modeling approach has been implemented in the prototype modeling system Spiff. This system provides interactive facilities for the specification of feature classes, and modeling facilities for the creation and manipulation of semantic feature models. The Cellular Model has been implemented using the Cellular Topology husk of the Acis Geometric Modeling kernel.

If you are interested in this thesis, we will be glad to send you a copy in book format (if still available); just mail us your post address.


download full thesis


Bidarra R (1999), Validity Maintenance in Semantic Feature Modeling, PhD Thesis, Delft University of Technology, ISBN 90-9012599-X.