Clustering of Composite Objects for CAD Databases

Composite objects are descriptions of collections of components which themselves can be collections of their subcomponents. Efficient storage structuring and fast retrieval of the composite objects are essential to computer aided design applications of database management systems. This paper introduces two clustering methods: a structural clustering and a function-propagation clustering, in which all components of a category are clustered into an equivalent object cluster. The structural clustering is an extention of the component aggregation method and the function-propagation clustering is an extention of the functional clustering method. These two clustering techniques partition the database into meaningful and application-oriented clusters. Several clustering methods are compared to demonstrate the performance enhancement when composite objects are organized according to our clustering methods.