Manufacturing industries have widely adopted the reuse of machine parts as a method to reduce costs and as a sustainable manufacturing practice. Identification of reusable features from the design of the parts and finding their similar features from the database is an important part of this process. Approaching Machining Feature Recognition (MFR) as a supervised problem is not feasible for real-time applications in industries. This is because different industries have their own data lakes to store the huge pool of 3D CAD models that include different combinations with respect to their requirements. Preparing a supervised corpus and training them to predict a large number of classes is a complex and tedious process to achieve. Here we proposed an approach that includes inductive transfer learning for geometrical feature extraction that works well in practical applications.

Website Link: [ Ссылка ]

arXiv Link: [ Ссылка ]

Github link to the code: [ Ссылка ]

свернуть