Reconstructing the topology of a CAD model - a discrete approach

We consider a problem which arises in the computer-aided design of machines , motor vehicles, and other technical devices: convert the data describing a surface model of a three-dimensional workpiece from a low-information for mat to a high-information format. In the worst case, the input format only contains a list of independent descriptions of the elementary pieces, where as the output format also contains the neighborhoods of these pieces-the to pology of the model. Reconstructing the topology is a crucial step in this process. This problem is ill-posed and hence cannot be solved fully automatically. I n fact, manual operations and previous format conversions make the data so "dirty" that human judgment is still required. Therefore, the aim is to app roximate the topology as closely as possible to reduce the effort needed to correct the result afterward. In this paper we present a discrete approach to this problem and report the results of a computational study.