Complex Parts Partitioning by Elementary Geometric Objects Clustering in CAD/CAM Process

Abstract

Complex shaped parts are omnipresent in the industry: aeronautics, molds, etc. CAD/CAM software is used in their production process. To simulate this process, three geometric representation models can be considered: workpiece model, tool model and swept volume model which can be discrete or continuous. The use of discrete models is simple to implement compared to continuous models but requires a tight discretization to represent faithfully theoretical shapes. This is necessary for a high accuracy but increases processing time. To overcome this computing constrains, the main contribution of this work consists in partitioning the part space into regions by pooling the geometric elements (triangles, Dexels) into sets while introducing for the first time in this context two methods knows as Fast Global K-means (FGK) and Incremental K-means (IK), that arise from Basic K-means (BK) method presented in previous works. Another key point is developed; it consists on establishing a pretreatment on the Triple Dexels model in order to be able to use BK, IK and FGK methods in the material removal simulation process. A comparative study has been established to evaluate the performances of these methods. FGK demonstrated better performances in terms of computational time, cluster uniformity and reduction of regions overlapping areas, compared to BK and IK methods.