Abstract
Nowadays, 3D CAD systems (Computer Aided Design) are extremely important in many sectors related to the modern industry. The wide range of modelling software and their increasingly advanced functionalities have considerably extended the uses of CAD models. Indexing and searching in huge databases for the purpose of reusing a 3D CAD model is still a big challenge for both the industrial and for scientific research fields. This highlights the need for an effective approach that provides accurate results within an acceptable time-frame and that takes into account the re-usability of the model regardless of its modelling software. From this perspective, this paper proposes a new data-intensive approach for the detection of reusable CAD models, similar to a reference one. The proposed method is based on the probabilistic graphical model Bayesian network. The suggested approach is mainly based on two phases: (1) first, the construction of a Bayesian network of models (MBN). The developed network is based on an index of association calculated according to the common reusable sub-parts between the candidate models. (2) Then, the detection of reusable models (through the indirect similarity extracted from the built MBN) based on the graph structure and the probabilistic inference.