Abstract: The key problem of CAD/CAE model transformation lies in how to decompose models to the simplest elements. Such element owns similar gridding property which can weigh computing error and time. And for this reason this paper proposed a feature recognition algorithm based on graph decomposition and decomposed the attributed adjacent graphs to many connectivity components, and then formed the volumetric feature in terms of these connectivity components. This algorithm is not limited to the feature type; it would return a desired result of model decomposition only if the resolvability of the vertex is controlled reasonably. And meanwhile, by using this algorithm we could acquire entity characteristics and remove or replace feature in the level of characteristics, which can simplify the model conveniently.

Key words: computer application, feature recognition, volumetric feature, graph decomposition