Live Product Line Engineering Using Density-Based Clustering of CAD Models

Manufacturers of special-purpose machines are faced with the challenge of having to efficiently engineer systems tailored to individual customer requirements, yet sharing the same base technology. One popular approach is that of “product line engineering” (PLE), the general idea being to maintain product lines, i.e., families of similar systems with variations in features, and to reuse components and assemblies wherever possible. We propose a live product line approach wherein variant information for components and assemblies is dynamically created and alternatives are proposed to the designer in the Computer Aided Design (CAD) system. To this end, we apply unsupervised learning techniques, in particular density-based clustering, to grouping CAD models of machine components. We demonstrate the feasibility using a data set from a special-purpose machine manufacturer and show that the approach can provide benefits in machine design.