• Geometric_and_Feature_Modelling
• Completed_Projects

Completed Projects

Assembly features in modelling and planning

Features have been introduced in modelling and planning for manufacturing of parts. In this project it has been shown that the feature concept is also useful in assembly modelling and planning.

For modelling and planning of both single parts and assemblies, an integrated object-oriented product model has been developed. For specific assembly-related information, assembly features are used. Handling features contain information for handling components, connection features information on connections between components.

A prototype modelling environment has been developed. A model can be manipulated using a geometric or a graph-oriented user interface. The model provides possibilities for component-driven and relation-driven modelling. In the first method, one starts with complete components, and by adding relations to them the final assembly is created. In the second method, one starts with relations, and from these the components can be created.

The product model has been successfully verified within several analysis and planning modules, in particular stability analysis, grip planning, motion planning and assembly sequence planning. Algorithms have been developed that exploit the information available in the features.

It has been shown that feature-based product models for assembly can considerably help in both assembly modelling and planning, on the one hand by integrating single-part and assembly modelling, and on the other hand by integrating modelling and planning.

Constraint-based feature validation

A constraint-based feature validation scheme has been developed that provides the means to specify and maintain the validity of features in multiple views. For feature validity specification, various constraint types of different domains have been identified: geometric, algebraic and topologic. A scheme for persistent feature validity maintenance has been developed, which is based on constraint management.

The constraint management is separated into two levels: constraint storage and constraint satisfaction. Firstly, the constraint management module maintains in a constraint graph the set of constraints that result from modelling actions. Secondly, it ensures satisfaction of all constraints by deploying several constraint solvers that are each specialized for a particular domain. The module maintains a coupling between the two levels, and it propagates solving results between the constraint solvers. The advantages of the two-level separation are that the interdependence of the constraints is handled by the constraint management module, while the constraints are satisfied efficiently by the solvers.

A geometric constraint solver has been developed based on extended 3D degrees of freedom analysis. Newly developed extensions include degrees of freedom ignorance for locus intersections, and restructuring of constraint loops. With these extensions, powerful 3D geometric constraint solving is achieved. For the analysis of over- and underconstrained geometric models, the concept of dependency graph has been developed, to identify the set of involved constraints.

Algebraic constraint solving is done with the SkyBlue solver; its constraint priority scheme is used to propagate feature parameter changes in such a way that the model is minimally disturbed.

Topologic constraints are checked by querying a cellular geometry representation; it allows detection of violations of topologic feature properties caused by feature interactions.

Product modifications are propagated between the views by link constraints that provide a generic coupling of the feature models of the views. A view priority scheme is used to handle conflicts between constraints of different views.

Feature conversion for concurrent engineering

Today's development of industrial products faces high demands. Products must be developed fast, and must be of good quality. For product developers, this implies integration of traditionally separated product development phases, i.e. concurrent engineering.

For concurrent engineering, functional product information is required. Design by features enables to model a product with both geometric and functional properties. Using feature validation, the intended functional product properties are maintained under model modifications.

Hence, the ideal product development environment integrates concurrent engineering and feature modelling. Each engineering discipline is represented by a so-called view. Each view contains features relevant to the specific discipline. Communication between different views requires feature conversion, which translates modifications to a feature model in one view to another.

For concurrent engineering using feature modelling, a product model has been developed. It consists of a shared central geometry and constraint representation, and a feature model for each view. The product geometry representation is a cellular data structure to which feature instances refer.

Two types of feature conversion have been developed that propagate feature model geometry and topology changes between views. Geometry changes are propagated using automatically derived geometric constraints. Topology changes are propagated using generic view specifications, the cellular model, and a newly developed feature recognition technique.

This new feature recognition technique is a generic and hierarchical technique which is called top-down feature recognition. At the top level, it prescribes feature classes in a specified order, ensuring meaningful feature interpretations. At the lower level, instances of a feature class are recognized based on geometric reasoning; it has the advantage that it can handle intersecting features.

This project has been supported by the Netherlands Computing Science Reserach Foundation (SION), with financial support from the Netherlands Organization for Scientific Research (NWO).

Validity maintenance in semantic feature modelling

Current feature-based modelling systems are still too much tied to methods and techniques of conventional geometric modellers. They strongly rely on a history-based approach of the modelling process and therefore do not offer the level of assistance that might be expected from feature technology. Among other pitfalls, the specification and maintenance of feature semantics is not well covered by such systems, and the semantics of modelling operations not well defined.

A complete solution to these problems poses a number of requirements at various levels of a feature modeller. Among them, the following goals are covered within the scope of this research work.

Feature libraries should include declarative specifications of feature classes, each of which contains a full description of the specific semantics desired for its feature instances. The feature model should be fully specified as a set of feature instances, related by their mutual positioning constraints.

The semantics of all modelling operations should be clear and unambiguous, through the consistent use of references to model features and their elements, instead of to geometric model entities.

Each modelling operation should be monitored, in order to assess the conformity of each feature in the model with its validity criteria. In particular feature interactions, caused by intersections of two or more features in a way that affects their functional or technological semantics, should be handled. Additionally, every validity violation should not only be detected but also be documented, reported to the user, possibly with context-dependent hints, and corrected.

As a consequence of the former, the structure of the model, as well as the interpretation of its evaluated geometry, should be completely and unambiguously determined without invoking any model history considerations. Additionally, the re-evaluation of the geometric model after each successful operation should be limited in scope, in order to keep its complexity independent of the number of features present in the model.

Multiple-view feature modelling with model adjustment

Multiple-view feature modelling is a recently introduced approach to product development, which combines concurrent engineering and feature modelling. It supports applications from various phases of product development, by providing an own interpretation of, or view on, a product for each of these applications. Each view has its own feature model of the product. The approach can lead to a higher quality of products in less time, which is one of the most important goals of contemporary product development.

Current approaches to multiple-view feature modelling still have at least three major shortcomings. First, they focus on the later product development phases, in which the geometry of the product has to be fully specified. Second, they deal with single parts only, whereas real product rarely consists of a single part. Third, they discard the possibility that a feature model of a product for some view is invalid or that a consistent feature model in another view cannot be created.

A new approach to multiple-view feature modelling has been developed that overcomes these shortcomings. It supports high-level product design and design of products with multiple parts, in particular conceptual and assembly design. Further, it supports automatic adjustment of form feature models that are invalid or for which no consistent form feature model in another view can be created.

The conceptual design view allows the designer to determine the configuration of a product by specifying components, which are to be implemented by one or more parts, and interfaces between them, which are to be implemented by a connection.

The assembly design view focuses on connection design, and allows the designer to specify the type of connection between components and the geometry for the connection on the components.

Consistency maintenance integrates all views with each other, by ensuring that their feature models represent the same product or part of it, i.e. that their feature models are consistent.

Automatic model adjustment is able to automatically adjust the model of a product in case the feature model of one of its views has become invalid, or a consistent feature model in another view cannot be created.

Collaborative feature modeling

Collaborative Modeling Systems are distributed, multiple-user systems that provide a team of geographically dispersed designers with the appropriate modeling tools for working together, concurrently and synchronously, on the design of a product. Such tools are especially helpful in current multi-disciplinary design teams, where different members have dissimilar views of the product according to their life-cycle activity (e.g., design, manufacturing planning, assembly planning, inspection planning). Effectively improving the collaboration within such teams can significantly reduce product cost and time to market. We believe that a multiple-view feature modeling system is the ideal basis for a collaborative modeling system.

Our research in collaborative feature modeling is currently focused, among other things, on the following crucial issues:

• concurrency, which deals with the problems of simultaneous access and manipulation of model data by different users;
• synchronization, which deals with the propagation of evolving data among system users, in order to keep their model data consistent;
• interactivity, which deals with visualization and user interaction facilities, aimed at guaranteeing an effective collaboration among system users.

Techniques being investigated within this project are implemented and tested in the collaborative feature modeling prototype system webSPIFF. This system presents a web-based client-server approach, where the server coordinates the collaborative session, maintains the shared model, and provides all functionality that cannot, or should not, be implemented on the client. The clients perform operations locally as much as possible, and only high-level semantic messages, as well as a limited amount of information necessary for updating the client data, will be sent over the network. A demo version of webSPIFF is available online at www.webSPIFF.org for all users interested in experimenting with it.